HK1203678B - Methods and systems for generating green score using data and sentiment analysis - Google Patents

Description

Method and system for generating green scores using data and sentiment analysis

Technical Field

The present invention relates generally to financial services and to mining information from traditional news sources and new/social media sources and other content sources to identify sentiment and predict behavior for pricing and recommendations. More specifically, the present invention provides intelligent analytics that enable the measurement and/or scoring of the "greenness" of companies and associated risk areas as perceived by traditional and new media and/or used to generate a composite "environmental" index, as well as predictive corporate valuation behavior. The present invention provides a dynamic tool that leverages machine learning capabilities, news sentiment expertise, and intelligent analytics to provide services for benchmarking environmental and sustainability sentiment of private and publicly traded companies.

Background Art

With the advent of the printing press, typesetting, typewriters, computer-enabled word processing, and mass data storage, the amount of information generated by humanity has increased dramatically and at an ever-increasing rate. Recently, less formal content sources, including "social media," have become increasingly prevalent. As opposed to traditional media, which is essentially passive (content is read), social media is more interactive, immediate, and often results in faster responses or reaction times. As a result, or as a result of the growth and diversification of information sources, there is a continuous and growing need to collect and store, identify, track, categorize, and catalog, as well as process this growing ocean of information/content and deliver value-added services to promote the intelligent use of data and predictive patterns derived from such information. With the development, widespread deployment, and accessibility of high-speed networks such as the Internet, there is a growing need to appropriately and efficiently process the ever-increasing amount of content available on such networks to aid decision-making. In particular, there is a need to quickly process information related to current events to enable informed decisions based on the impact of current events or related sentiment, and to consider the potential impact such events and sentiment may have on the prices of traded securities or other offerings. The widespread availability and access to blogs, wikis, forums, chat rooms, and social media has enabled a growing number of information recipients to express opinions about people, companies, governments, and commercial products. Virtually instant and simultaneous access to information can increase the correlation between events and stock prices.

In many fields and industries, including the financial services industry, there are content and enhanced experience providers such as The Thomson Reuters Corporation, Wall Street Journal, Dow Jones News Service, Bloomberg, Financial News, Financial Times, News Corporation, Zawya, and the New York Times. Such providers identify, collect, analyze, and process key data for use in generating content such as reports and articles for consumption by professionals and others involved in the corresponding industries (e.g., financial advisors and investors). Using one method of content delivery, these financial news services provide both real-time and archived financial news feeds that include articles and other reports on recent events of interest to investors. Many of these articles and reports, both actual and potential events, can have a measurable impact on the trading stock prices associated with publicly traded companies. While this article often discusses publicly traded stocks (e.g., those traded on markets such as the NMASDAQ and the New York Stock Exchange), the present invention is not limited to stocks and includes application to other forms of investments and securities. Professionals and providers across a wide range of industries are continually seeking ways to enhance the content, data, and services they offer to subscribers, clients, and other customers, and to differentiate themselves from the competition. Such providers are committed to creating and delivering enhanced tools, including search and ranking tools, to enable customers to process information more efficiently and effectively and make informed decisions.

Advances in technologies including database mining and management, search engines, language recognition, and modeling have provided increasingly sophisticated methods for searching and processing large amounts of data and documents (e.g., news articles, financial reports, blogs, SEC and other required corporate disclosures, legal decisions, decrees, laws, and regulatory databases) that may affect business performance and, therefore, the prices associated with the stocks, securities, or funds comprising such holdings. Investment and other financial professionals, as well as other users, increasingly rely on mathematical models and algorithms to make professional and business decisions. Particularly in the investment field, systems that provide faster access to and processing of (accurate) news and other information related to corporate performance would be highly valuable tools for professionals and would lead to more informed and successful decision making.

Many financial service providers use "news analytics" or "news analytics" to provide enhanced services to subscribers and customers. "News analytics" or "news analytics" refers to a broad field encompassing and involving information retrieval, machine learning, statistical learning theory, network theory, and collaborative filtering. News analytics encompasses a set of techniques, formulas, statistics, and related tools and metrics used to comprehend, summarize, categorize, and otherwise analyze information sources (often publicly available "news" information). An exemplary use of news analytics is a system that comprehends (i.e., reads and categorizes) financial information to determine the market impact associated with such information while normalizing the data for other effects. News analytics refers to measuring and analyzing various qualitative and quantitative attributes of textual news reports, such as those found in formal, text-based articles and less formal delivery methods such as blogs and other online media. More specifically, the present invention focuses on analysis within the context of electronic content. Attributes include sentiment, relevance, and novelty. Expressing or representing news reports as "numbers" or other data points enables the system to transform traditional information representations into mathematical and statistical representations that are more easily analyzed. News analysis techniques and metrics can be used in a financial context, and more particularly in the context of past and predictive investment performance.

News analytics systems can be used to measure and predict the following: earnings, stock valuations, market volatility; the impact of news; the relationship between news and message board information; the correlation of risk-related terms in annual reports to predict negative returns; sentiment; the impact of news coverage on stock returns; and determining the impact of optimism and pessimism in news on earnings. News analytics can be viewed at three levels or layers: text, content, and context. Much effort has focused on the first layer—text—where text-based engines/applications process the raw textual components of news, such as words, phrases, and document titles. Text can be converted or utilized into additional information, and irrelevant text can be discarded, resulting in a concentration of information with higher relevance/usefulness. The second layer (content) represents the richness of the text, which can be further exploited by analytics to impart higher meaning and significance, such as quality and truthfulness. Text can be categorized as either "factual" or "opinion" expressions. The third layer (context) of news analytics refers to the connectivity or relationships between information items. Context can also refer to the network relationships within news. For example, Das and Sisk (2005) examined the social networks of message board posts to determine whether portfolio rules could be formed based on the network connections between stocks.

After processing news reports based on text, content, and context, investors and those involved in financial services want to understand how this large amount of information (even processed information) relates to the likely movement of a company's stock price. A commonly used term and measurement related to company risk is "alpha." As used in this application, "alpha" represents a measure of performance on a risk-adjusted basis. For example, alpha takes into account the volatility (i.e., price risk) of an instrument, such as a stock, bond, mutual fund, etc., and compares risk-adjusted performance to another performance measure (e.g., a benchmark or other index). The return of an investment vehicle (e.g., a mutual fund) compared to the return of a benchmark (e.g., an index) is the investment vehicle's alpha. Additionally, alpha can refer to the abnormal return of a security or portfolio that exceeds what would be predicted by an equilibrium model (such as the capital asset pricing model). Alpha is one of five widely considered technical risk ratios. In addition to alpha, other statistical measures of technical risk factors used in modern portfolio theory include beta, standard deviation, R-squared, and the Sharpe ratio. These statistical risk indicators are used by investment firms to determine the risk-return profile of stocks, bonds, or other investment vehicles based on securities, such as mutual funds. For example, in the case of a mutual fund, an alpha of positive or negative 1.0 means that the mutual fund outperformed its benchmark index by positive or negative 1%, respectively. Accordingly, if a Capital Asset Pricing Model analysis estimates that a portfolio should return 10% based on its riskiness, and the portfolio actually returns 15%, then the portfolio's alpha would be positive 5%, representing an excess return over and above what the model analysis predicted.

Particularly as it relates to the present invention, increasing pressure from government regulators and an increasingly "green"-conscious public has led to a growing demand from interested parties (e.g., the investment community and others in the financial services industry) for new tools to assess the degree of "greenness" (or green score or factor) and/or environmental compliance of companies/investments, as well as key areas for managing risk exposure. Investment firms and managers focused on green/environmental investments need solutions that provide information related to a company's greenness and/or environmental compliance, as well as tools for evaluating them. As used herein, "greenness" refers to a company's products, manufacturing, distribution, packaging, or other corporate practices that relate to the environmental impact of the company and its products. For example, a product's green score may take into account the use of recycled materials, the amount of energy required to operate the product, the electromagnetic effects of the product, and the amount of harmful emissions or pollution emitted by the product. Countries and regions have enacted legislation, regulations, certifications, standards, and other requirements related to the operation of products and the disposal, recycling, and treatment of such products (e.g., RoHS (EU)). Certain manufacturing processes and materials have been found to have harmful environmental impacts and are subject to restrictions or regulations. Certain practices have been found to promote or address environmental sustainability. Companies may go "paperless" in their operations and incorporate environmentally friendly materials and systems into their facilities. Allowing employees to work from home can contribute to reduced commuting burdens, reduced consumption of natural resources, and reduced harmful emissions.

Beyond investment considerations, businesses are increasingly aware of and focused on integrating green investments with governance, risk, and compliance (GRC), corporate social responsibility (CSR), and environmental, social, and governance (ESG) initiatives. Solutions are needed to help these companies evaluate and track the effectiveness and performance of their green investments and efforts. Tools are also needed to help manage market and reputational risks arising from negative trends and demonstrate a certain level of compliance with green/social standards. Furthermore, regulatory agencies and other institutions require solutions that help them identify and manage potential hotspots, such as topics or geographic areas of environmental concern, when debating, proposing, and enacting influential green legislation.

Green-related actions can have significant impacts on a variety of issues, directly and indirectly affecting businesses, market indices, and investors in equities, bonds, and more. A recent example of a green-related event impacting valuations and behavior is the explosion of an offshore drilling platform off the Louisiana coast in the Gulf of Mexico, resulting in the oil spill disaster. This event significantly impacted the financial performance of several entities, including publicly traded British Petroleum ("BP"). News of the disaster had the immediate effect of causing BP's common stock to plummet on the day of the disaster and in the days that followed. In addition to asset losses, oil cleanup costs, and compensation claims from those adversely affected by the spill, BP also suffered the resulting political and social fallout. The Exxon Valdez oil tanker grounding and resulting spill is another such example. While many organizations track such events and may maintain company scorecards showing relative performance, there is no system that effectively monitors events and provides investors with simultaneous information on how such events may affect corporate performance (e.g., stock prices).

The "green analytics" space is rich and rapidly growing, with investment firms and managers driving much of the growth in green analytics and holding the highest projected demand. Existing products within the green analytics space generally fall into three categories: ESG risk solutions, thematic indexes and benchmarks, and reputation monitoring. One provider within this space is RiskMetrics/KLD, which specializes in web-based research services, thematic indexes, and carbon analytics. Financial services firms offer ESG products through indexes and web-based research platforms. Societe General, for example, offers thematic indexes covering a variety of issues from human rights to CSR. Other players, such as FTSE, Dow Jones, and Calvet Investments, offer environmental indexes that investors can use for benchmarking and portfolio construction. Within the reputation monitoring space, companies such as RepRisk and Factiva Insight offer web-deployed tools that can be broadly intelligence-based or focused, such as brand risk, as it relates to environmental issues. Third-party sources can be used to visually process and deploy analyst sentiment via the web, allowing customers to monitor negative green news by company and industry.

All of these efforts have drawbacks, including the inherent redundancy that obscures green-oriented products. These efforts to measure a company's greenness are flawed in that they use the same sources from which each metric is derived (i.e., third-party research, corporate filings, regulations). Furthermore, the assessments are performed by analysts and are highly dependent on the timeliness of public filings and secondary research, similar to the dilemma faced by credit rating agencies competing with real-time credit default swap curves.

Currently, clients are faced with a market of products offering essentially the same human-driven research tools, albeit with varying deployment methods and visualizations. Asset managers serving green-minded retail and institutional investors may find it difficult to leverage these tools to fulfill their mandates to invest in green companies and, perhaps more importantly, to communicate the value of these investments to their clients. A recent study conducted by the University of Zurich highlights this dilemma. Using ESG data from RepRisk, the study compared the sustainability of green funds with that of conventional equity funds.

These tools are largely driven by the same sources and basic analysis means they can produce similar results that do not fully capture the perceptions associated with being green. Arguably, these tools ignore underlying trends from non-traditional sources that add enormous value to decision making.

The same idea easily applies to businesses and regulators. Faced with the need to monitor their brands and manage reputational risks stemming from poor CSR performance and bad PR, businesses need a tool that is regularly updated and systematically leverages the vast array of new media. More importantly, they need a tool that captures the perceptual elements missing from other offerings. At the same time, regulators are now tasked with managing environmental hotspots not only at the industry level but also at the corporate level, particularly when the companies in question receive public funding.

What is needed is a system that can automatically process or "read" available news reports, filings, news/social media, and other content and quickly interpret it to gain a better understanding of the environmental impact of an entity (private or public). There is also a need to create and apply predictive models to anticipate the behavior of stock prices and other investment vehicles based on the entity's environmental impact before they actually change. There is a need to use and leverage traditional and especially new media sources and trends, as well as to meet consumer demand for advanced analytics related to corporate performance, price behavior, investments, and reputational awareness, to provide a sentiment-based solution that extends the scope of conventional tools to include social media and online news.

Summary of the Invention

The present invention uses and utilizes new media resources and trends to meet customer demand for advanced analytics related to ESG mandates, green investments, and reputation awareness. Social media has an increasing influence on environmental issues. With the publication of carbon legislation and the commercialization of a global culture toward "greenness," the impact of new media on environmental and social governance will increase over time. The present invention provides a green sentiment solution in its embodiments that expands the scope of conventional tools to include social media and online news to generate and present enhanced tools, content, and solutions. The present invention provides an indication of an entity's environmental behavior through a simple score that can be negative or positive and evolves over time. Smart analytics allow customers to measure the "greenness" of a company as perceived by conventional and new media. The solution aggregates private and public content from multiple sources, including social media content. The classification method is tuned to understand topics, text, phrases, sentences, comments, and other content as having or not having green or environmental meanings. The results can take the form of one or more of the following: a green score, a composite environmental or green index, and a green company certification or classification.

In one implementation, the present invention provides a news/media analysis system (NMAS) and related methods that are adapted to automatically process and "read" news stories and content from blogs, twitter, and other social media sources in as close to real time as possible. The present invention employs quantitative analysis, techniques, or mathematics in conjunction with computer science to derive green scores, green certifications, and/or model the value of financial securities, including generating a composite environmental index. The present invention provides a system for automatically processing or "reading" news stories, filings, new/social media, and other content and for applying predictive models to said content to anticipate the behavior of stock prices and other investment vehicles. NMAS leverages traditional and especially new media resources to provide a sentiment-based solution that expands the scope of conventional tools to include social media and online news.

In addition to, and in some ways a replacement for, traditional media sources and means of delivery, "social media" has added a new level of information sharing and gathering that goes far beyond conventional media forms. Unconstrained by traditional models and workflows, blogs and other social media forms have become highly accessible and widespread sources of real-time news and updates. On the investment front, startups like Seeking Alpha, as well as traditional financial news providers, are moving into the blogosphere and social media at an exponential rate. Blogs and other new media have become the most important source of investment advice, and for some, surpass traditional sources. "Social media" or social network sources refer to non-traditional, often less formal forms of content delivery, and include interactive user- or crowd-sourced data and content. Examples of social media include: news websites (reuters.com, bloomberg.com, etc.); online forums (livegreenforum.com); government agency websites (epa.gov); websites for academic institutions and political parties (mcgill.ca/mse, www.democrats.org, etc.); online magazines (emagazine.com/); blogs (Blogger, ExpressionEngine, LiveJournal, Open Diary, TypePad, Vox, WordPress, Xanga, etc.); microblogging sites (Twitter, FMyLife, Foursquare, Jaiku, Plurk, Posterous, Tumblr, Qaiku, Google Buzz, Identi.ca Nasza-Klasa.pl, etc.); social and professional networking sites (Facebook, MySpace, ASmallWorld, Bebo, Cyworld, Diaspora, Hi5, Hyves, LinkedIn, MySpace, Ning, Orkut, Plaxo, Tagged, XING, IRC, Yammer, etc.); online advocacy and fundraising sites (Greenpeace, Causes, Kickstarter); information aggregators (Netvibes, Twine, etc.); and Twitter.

In one embodiment, a private investor sensitive to the environmental behavior of an entity can use the present invention to monitor and collect information from social media that would not otherwise be available, or at least delayed, when monitoring traditional "mainstream" or conventional media. With the increasing adoption of new social media, such sources are becoming increasingly "mainstream." Furthermore, the present invention can be used to aggregate content from multiple social media content producers to verify, validate, or otherwise enhance the collected information.

NMAS can include sentiment processing to process news/media information and assign a "sentiment score" to news/media items related to one or more companies. The score can be derived from the text and metadata from the news/media, and predefined or learned lexicon-based and/or sentiment models can be applied to the processed text/metadata. NMAS can include a training or learning module that analyzes past news/media and the resulting related stock price responses based on certain events to build models for predicting stock behavior given certain types of news or events, including those related to green or environmental events, certifications, legislation, etc.

In one approach, the present invention can be used to process traditional and new media content sources as sources for determining or representing "greenness" or "alpha" in the context of a composite environmental index. In an exemplary implementation, an NMAS operated by a traditional financial services company can apply internal textual sources and external sources to predictive models to derive expected market-related behavior. Hard facts and sentiment are considered as factors driving green scores and/or composite environmental indices. NMAS news/media sentiment analysis and green scores enhance investment and trading strategies and lead to informed trading and investment decisions.

Furthermore, the present invention can be used to generate a classification system for environmentally conscious or environmentally friendly companies, serving as a classification system for green investing. The present invention can be used to classify or certify a company as "green compliant," as well as to create a "green sentiment index" comprised of companies that have received green certifications. Green indices have the potential to attract investors interested in promoting environmentally responsible businesses.

Unlike other approaches that rely on periodic research conducted by analysts, the present invention continuously processes media feeds and generates information and data streams that capture daily trends, along with the added value of portals and intelligent alerts that allow users (e.g., customers) access to a range of content. As green or environmental news and social media content grows, media services companies will leverage products and services across a broad offering platform, such as Thomson Reuters Markets. The present invention enables companies to connect offerings across sectors and accelerate market penetration in the green analytics space.

The present invention can be used to track "green" sentiment over time to provide analysis of company-related news/media commentary, as well as tools and analytics to guide trading and investment decisions based on green or environmental issues. The present invention can be powered by natural language processing using linguistic techniques. The present invention provides quantitative "green" strategies that support human decision-making, risk management, and asset allocation. The present invention can be used in market making, in portfolio management to improve asset allocation decisions by benchmarking portfolio sentiment and calculating industry weightings, for fundamental analysis to forecast stock, industry, and market outlooks, in risk management to better understand unusual risks to the portfolio and to develop potential sentiment protections, and to track and benchmark public perception and media coverage, as well as competitor performance.

In a first embodiment, the present invention provides a computer-implemented method comprising: identifying an entity to be assigned a green score; calculating a green score based on a set of social media information; and transmitting the green score. The set of social media information may be associated with the identified entity, the entity being associated with a traded security, the set of social media information representing green information related to the entity, and the green score representing a green attribute of the entity. The method may also include modifying the green score over time based on additional sets of social media information, and transmitting the modified green score. Calculating the green score may include determining a sentiment score based at least in part on the set of social media information, and wherein the green score represents a green sentiment score. The method may also include: authenticating the entity as green-compliant based at least in part on the green score; performing steps (a)-(c) for a second identified entity results in a second green score, and wherein transmitting the green score includes transmitting a data feed including data reflecting the green score and the second green score; generating an alert signal related to the entity based at least in part on the value of the calculated green score; and/or identifying text in a content set received from a social media collection that is identified as representing the entity, and further including extracting additional text from the content set that is identified as representing sentiment related to the greenness of the entity. The identified text in the content set may include one or more of: identifying embedded metadata or other descriptors; processing text, words, phrases; applying natural language linguistic analysis; applying Bayesian techniques. The method may further include aggregating a set of content from a plurality of sources, the plurality of sources including at least one social media source and at least one additional source from the group consisting of: news websites (reuters.com, bloomberg.com, etc.); online forums (livegreenforum.com); websites of government agencies (epa.gov); websites of academic institutions, political parties (mcgill.ca/mse, www.democrats.org); online magazine websites (emagazine.com); blog websites (Blogger, ExpressionEngine, LiveJournal, Open Diary, TypePad, Vox, WordPress, Xanga); microblogging websites (Twitter, FMyLife, Foursquare, Jaiku, Plurk, Posterous, Tumblr, Qaiku, Google Buzz, Identi.ca, Nasza-Klasa.pl); social and professional networking sites (facebook, myspace, ASmallWorld, Bebo, Cyworld, Diaspora, Hi5, Hyves, LinkedIn, MySpace, Ning, Orkut, Plaxo, Tagged, XING, IRC, Yammer); online advocacy and fundraising sites (Greenpeace, Causes, Kickstarter); information aggregators (Netvibes, Twine, etc.); Facebook; and Twitter, the content set comprising the social media information set. The method may also include: identifying a set of companies to be associated with a composite index, the set of companies including the entity and associated with the set of securities; generating a composite context index for the set of securities based at least in part on the set of social media information; and transmitting a signal associated with the composite context index. The composite environmental index is determined at least in part based on a set of green scores associated with the set of companies, and the green scores can be generated and/or derived in real time based on one or more of the following positive criteria: compliance or certification related to product or manufacturing environments; energy efficiency; corporate practices that promote environmental stewardship, consumer protection, human rights, and diversity; businesses/products involved in green technology, energy-efficient technology, alternative fuel technology, and renewable resource technology; and/or based on one or more of the following negative criteria: businesses involved in alcohol, tobacco, gambling, weapons, and/or the military; and businesses that do not comply with environmental standards. The entity is associated with securities traded in the market, and further includes applying a predictive model to derive predicted behavior associated with the securities. The method may also include generating an expression of the predicted behavior and/or a recommended action to be taken based on the predicted behavior. The recommended action relates to a trading decision related to the security and is one of the group consisting of buy, sell, or hold. The set of social media information is identified based on time value. The method may also include: generating a risk signal representing a potential risk; providing a risk indicator pattern set on a computing device; and identifying a potential risk set associated with the entity within the set of social media information using a risk identification algorithm based at least in part on the risk indicator pattern set; comparing the potential risk set to the risk indicator pattern to obtain a prerequisite risk set; generating a signal representing the prerequisite risk set; and storing the signal representing the prerequisite risk set in an electronic memory. The method may also include: repeating steps (a)-(c) for an entity set resulting in a green score calculated for each entity in the entity set; identifying a green classification; and selecting entities from the entity set for inclusion in the classification based at least in part on the corresponding entity green score. The classification may involve certifying a company as green compliant, and wherein each of the selected entities is certified as green compliant.

In a second embodiment, the present invention provides a computer-based system comprising: a processor adapted to execute code; a memory for storing the executable code; adapted to receive an input of a set of social media information associated with an entity to which a green score is to be assigned; a green score module adapted to be executed by the processor, and the green score module including code executable by the processor to calculate a green score based on the set of social media information; and an output adapted to transmit a signal associated with the green score.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to facilitate a full understanding of the present invention, reference is now made to the accompanying drawings, in which like elements are referred to using like reference numerals. These drawings should not be construed as limiting the present invention, but are intended to be illustrative and for reference purposes only.

FIG1 is a first schematic diagram illustrating an exemplary computer-based system for implementing the present invention;

FIG2 is a second schematic diagram illustrating an exemplary computer-based system for implementing the present invention;

FIG3 is a search flow chart illustrating an exemplary method for implementing the present invention;

FIG4 is a flow chart illustrating database and document processing, sentiment, and green scoring using predictive modeling as input and output of a system employing the present invention;

5 is a flow chart illustrating an exemplary method for generating emotions for use in green scoring in conjunction with the present invention;

FIG6 is a diagram showing the expression of a green group in the form of a website incorporating the present invention;

FIG. 7 shows exemplary forms of output or services incorporating the present invention; and

8-16 are examples of risk mining techniques for use in implementing the present invention.

DETAILED DESCRIPTION

The present invention will now be described in more detail with reference to the exemplary embodiments as shown in the accompanying drawings. Although the present invention is described herein with reference to exemplary embodiments, it should be understood that the present invention is not limited to such exemplary embodiments. Those skilled in the art who can use the teachings herein will recognize additional implementations, modifications and embodiments and other applications for using the present invention, which are fully contemplated within the scope of the invention disclosed and claimed herein, and with respect to which the present invention may have significant utility.

The present invention uses and leverages new media resources and trends to meet customer needs for advanced analytics related to CSR, ESG mandates, green investing, and reputation awareness. The present invention, in its various embodiments, provides a green sentiment solution that expands the scope of conventional tools to include social media and online news to generate and present enhanced tools, content, and solutions. The present invention includes intelligent analytics that analyze conventional and new media to measure the "greenness" of a company and the resulting score that represents the entity's environmental behavior. The greenness score can be a simple score that can be negative or positive and can evolve over time. The present invention aggregates private and public content from multiple sources, including social media or web content, news, websites, and institutional newswires (e.g., Twitter, Facebook, websites, RSS). The taxonomy is tuned to understand topics, text, phrases, sentences, comments, and other content as having or not having green or environmental meaning.

The present invention may include emotion, feeling, and affect computing techniques to analyze text to identify human emotions related to green issues that impact a company's performance and predict further human responses, such as selling or buying company-related securities. Human emotions can be viewed as time-derived functions with a series of related causes and effects, or "effects and effects." For example, in a given situation, such as a person facing a potentially deadly conflict, a human emotion of fear can be expected to be followed by one or more alternative human responses, such as flight or defense. One or more expected future reactions to the situation can be represented using probabilistic values or relationships. Bayesian networks are often used to represent causal relationships. Additional data can be used to further refine or define the one or more probabilistic relationships. For example, if the person being threatened is armed, the probability of self-defense can be adjusted upward and the probability of flight can be adjusted downward. Similarly, if the person is cornered or otherwise limited in their means of escape, the probabilities can be adjusted. The present invention uses the detected human emotions to predict further human reactions, and does so on a collective basis. The system can then predict or predict human responses to the predicted emotions, such as generally selling a stock or selling a specific stock that is the subject of a negative press release. The present invention collects or uses or observes human emotions related to objects expressed on blogs, wikis, online forums, chat rooms, message boards and social media networks to detect "sentiment" related to green issues, such as statements by companies about using "green" or environmentally friendly raw materials or materials or practices. The present invention processes the collected information using the techniques discussed herein to derive a green score or rating based on the determined sentiment. The score can then also be used to recommend companies or alerts or otherwise identify companies for investment consideration. The present invention can also be used to generate a composite index of companies that meet selection criteria, such criteria being related to environmentally conscious or environmentally sensitive practices. In this way, investors, individuals, funds, etc. can use such scores, ratings or indexes as the basis for investment decisions.

In one implementation, referring to FIG1 , the present invention provides a News/Media Analysis System (NMAS) 100 adapted to automatically process and "read" news stories and content from blogs, Twitter, and other social media sources represented by a News/Media Corpus 110 in as close to real time as possible. Quantitative analysis, techniques, or mathematics incorporating computer science, such as a green scoring/composite module 124 and a sentiment processing module 125, are processed by a processor 121 of a server 120 to derive green scores, green certifications, and/or model the value of financial securities, including generating composite environments or green indexes. NMAS 100 automatically processes news stories, filings, new/social media, and other content and applies one or more models to the content to determine green scores and/or expected behavior of stock prices and other investment vehicles. NMAS 100 leverages traditional and, in particular, new media sources to provide a sentiment-based solution that extends the scope of conventional tools to include social media and online news.

NMAS 100 can receive content from the following exemplary new and social media sources as input via new media sources 1141, blogs 1142, and social media 1143 in the news/media corpus 110: news websites (reuters.com, bloomberg.com, etc.); online forums (livegreenforum.com); government agency websites (epa.gov); websites of academic institutions, political parties (mcgill.ca/mse, www.democrats.org, etc.); online magazine websites (emagazine.com/); blog websites (Blogger, ExpressionEngine, LiveJournal, Open Diary, TypePad, Vox, WordPress, Xanga, etc.); microblogging websites (Twitter, FMyLife, Foursquare, Jaiku, Plurk, Posterous, Tumblr, Qaiku, Google Buzz, Identi.ca, etc.); Nasza-Klasa.pl, etc.); social and professional networking sites (Facebook, MySpace, ASmallWorld, Bebo, Cyworld, Diaspora, Hi5, Hyves, LinkedIn, MySpace, Ning, Orkut, Plaxo, Tagged, XING, IRC, Yammer, etc.); online advocacy and fundraising sites (Greenpeace, Causes, Kickstarter); information aggregators (Netvibes, Twine, etc.); Facebook; and Twitter.

The NMAS 100 of FIG1 includes a sentiment processing module 125 adapted to process news/media information received as input via the news/media corpus 110 and assign a "sentiment score" to news/media items related to one or more companies. Sentiment and sentiment scores can be derived from computational linguistics and, for example, typically define or represent the tone of an article, blog post, social media comment, etc. as positive, negative, or neutral, using scores of +1, -1, and 0, respectively. The scores can be derived from text and/or metadata (either existing or newly assigned by the engine) from the news/media, and predefined or learned lexicon-based and/or sentiment models can be applied to the processed text/metadata. The NMAS 100 can also include a training or learning module 127 that analyzes past or archived news/media and the resulting associated stock price responses based on certain "facts" or events to build models for predicting stock behavior given certain types of news or events, including news or events related to green or environmental issues, campaigns, legislation, and the like.

In one approach, NMAS 100 can be used to process traditional and new media content sources 110 as sources for determining or representing "greenness" or "alpha" in the context of a composite environmental index. In an exemplary implementation, NMAS 100 is operated by a traditional financial services company (e.g., Thomson Reuters), where the primary database, internal 112, is an internal text source (e.g., TR News and TR Feeds). NMAS 100 applies this data to a green scoring module 124 and a sentiment processing module 125 and may include predictive models for deriving expected market-related behavior. For example, Thomson Reuters sources, as the primary internal database, may include legal sources (Westlaw), regulations (particularly SEC, dispute data, industry-specific, etc.), social media (with special metadata applied to make them useful), and news (Thomson Reuters News) and news-like sources, including financial news and reports. Furthermore, external sources 114, either freely available or subscription-based, may be used to supplement internal sources 112 as additional data points considered by the predictive models. Both hard facts (e.g., an oil well explosion resulting in direct financial losses (lost revenue, damages, etc.) and negative environmental impacts, resulting in a negative greenness score) and sentiment (e.g., the effects of quantitative fear, uncertainty, negative reputation, etc.) are considered factors driving green scores and/or composite environmental or green indexes. The results can be used to enhance investment and trading strategies (e.g., stocks and other equities, bonds, and commodities), enabling users to track and discover new opportunities and generate alpha. News/media sentiment analysis 125 can be combined with the green scoring module 124 to provide green scores to drive informed trading and investment decisions.

Furthermore, the NMAS 100 may include a green classification module 128, which is adapted to generate a classification system for environmentally conscious or environmentally friendly companies, serving as a classification system for green investing and can be used to create a composite environmental index. For example, a company currently assigned a RIC (Reuters Certification Code, a class of ticker codes used to identify financial certifications and indices) may be classified as "green compliant" (e.g., documented/maintained with a green score of a certain level and/or duration). In this manner, the present invention can be used to create a green RIC classification for trading purposes. For example, a "green sentiment index," such as one comprised of companies that have obtained green certifications or green RICs, may be generated and maintained. A green index has the potential to attract investors interested in promoting environmentally responsible businesses.

In one embodiment, the NMAS 100 may include a training or machine learning module 127 (such as ThomsonReuters' Machine Learning Capabilities and News Analytics) to derive insights from a broad corpus of environmental data, news, and social media to provide normalized green scores at the company (e.g., IBM) and index level (e.g., S&P 500). This historical database or corpus may be separate from or derived from the news/media corpus 110.

Preferably, the company's green score or index is calculated in near real time (e.g., approximately 150ms) and is used, for example, to develop an Alpha strategy for investment, monitor a company's green reputation, and identify changes in risk profiles at the company and industry level. Unlike other methods that rely on periodic research processed by analysts, the present invention receives and continuously processes media feeds other than traditional sources, such as WWW web and social media feeds. In one embodiment, the present invention generates information and data streams that capture daily trends and allow users (e.g., customers) to access portals and smart alerts for a range of content from, for example, related and unrelated products (e.g., other Thomson Reuters products). As green or environmentally related news and social media content increases, media service companies can utilize products and services across a broad supply platform, such as Thomson Reuters Markets. The present invention enables companies to connect supplies across divisions and accelerate market share penetration in the green analytics space.

For example, the green score criteria applied by the green scoring module 124 of the NMAS 100 may include: compliance or certification related to the product or manufacturing environment; energy efficiency; and corporate practices that promote environmental stewardship, consumer protection, human rights, and diversity. The green score criteria applied by the NMAS 100 may also include: positive attributes or scores for businesses/products involved in green technology, energy-efficient technology, alternative fuel technology, and renewable resource technology, as well as negative attributes or scores for businesses involved in alcohol, tobacco, gambling, weapons, and/or the military. Areas of concern recognized by the SRI industry can be summarized as environment, social justice, and corporate governance (ESG). Although described in terms of greenness and environmental compliance, the present invention may also be applied to creating a healthy lifestyle based on social goals and aspirations, or other categories for scoring companies.

NMAS 100 can be powered by natural language processing that utilizes linguistic techniques to process news/media data and content delivered to it. NMAS 100 analyzes company-related news/media commentary to track "green" sentiment over time. The quantitative "green" strategies provided by NMAS 100 can be used in market making, in portfolio management to improve asset allocation decisions by benchmarking portfolio sentiment and calculating industry weightings, in fundamental analysis to forecast stock, industry, and market outlooks, in risk management to better understand unusual risks to the portfolio and develop potential sentiment protections, and to track and benchmark public perception and media coverage and to do the same for competitors.

NMAS 100 can automatically analyze news content and generate trading (e.g., buy/hold/sell) signals and/or update green scores and/or composite environmental index information in near real time. As used herein, the term "near real time" means within one second. However, the broader the scope of data used in conjunction with the NMAS, the longer the response time may be. To shorten response time, a smaller window/amount of data/content can be considered. Furthermore, NMAS can be configured to maintain a rolling dataset so that it only updates existing scores and reports and, at any given moment, only processes ("reads" and scores and predicts) based on newly discovered, received, or published content from any source. NMAS scans and analyzes news and social media content about thousands of companies in near real time and feeds the results into quantitative strategies and predictive models. NMAS output can be used to power quantitative strategies across markets, asset classes, and all trading frequencies, support human decision making, and assist in risk management, investment, and asset allocation decisions.

Content may be received as input to NMAS 100 in any of a variety of ways and forms, and the present invention is independent of the nature of the input. Depending on the source of the information, NMAS will apply various techniques to gather information relevant to the green score. For example, if the source is internal or otherwise in a format recognized by NMAS, it may be possible to identify content related to a particular company or industry or index based on fields or tags in metadata associated with or identifying the document. If the source is external or not otherwise in a format easily understood by NMAS, natural language processing and other linguistic techniques may be employed to identify the company in the text and to which the statement relates. Additional such techniques can be used to identify text terms of potentially enhanced relevance, such as binning text across the following exemplary primary dimensions: "Author Sentiment" - a measure of how positive, negative, or neutral the tone of the article is regarding the item, specific to each company in the article; "Relevance" - how relevant or substantive the coverage is to a particular item; "Quantity" - how much news is happening about a particular company; "Uniqueness" - how fresh or repetitive the item is over different time periods; and Headline Analysis - indicating, among other things, special features such as broker action, pricing commentary, interviews, exclusives, and composite coverage. NMAS uses rich metadata, such as: company identifiers; topic codes - identifying the subject matter; stage of coverage - alert, article, update, etc.; and business industry and geographic classification codes; index references to similar articles. Metadata across multiple domains provides differentiated content for use by quantitative analysts and sophisticated algorithmic engines.

NMAS can utilize various and diverse text scoring and metadata types. The following are exemplary types for use with the present invention: Item Type - Alert, Article, Update, Correction; Item Genre - the classification of the report, i.e., interview, exclusive, composite report, etc.; Title - the alert or headline text; Relevance - 0-1.0; General Sentiment - 1, 0, -1; Positive, Neutral, Negative - which provide a more detailed indication of sentiment; First Mention Location - the location of the sentence where the item is first mentioned; Total Sentences - used for article length; Number of Companies - how many companies are tagged to the item; Number of Words/Logos - how many words/logos are mentioned about the company; Total Words/Logos - the number of words/logos in the news item. Total number of tags; Broker Action - indicates the broker action: upgrade, downgrade, hold, undefined, or whether it is the broker itself; Price/Market Commentary - used to tag items that describe pricing/market commentary; Item Count - how many items have been published about a company in different time periods; Link Count - indicates the degree of overlap from 12 hours to 7 days; Topic Code - which describes what the article is about, i.e. RCH = Research; RES = Results; RESF = Results Forecast; MRG = Mergers and Acquisitions, etc.; Other Companies - what other companies are tagged to the article; and other metadata - index ID, link reference, story chain, etc.

Figures 1-4 illustrate exemplary structural components and frameworks for carrying out the present invention and for providing an effective interface for user interaction with such computer and database-based systems. The following is a more detailed description of the implementation of the processes and features of the present invention, including a discussion of low-frequency work on news sentiment, and general exploratory data analysis on equities (including volatility and direction) and commodities. In an exemplary scenario, which is not intended to limit the present invention but merely to aid illustration, the following describes how news metadata is related to prices and discusses the short-term relationship between news and prices. The exemplary discussion examines four equity markets (US, UK, Japan, and Hong Kong) and four commodities (crude oil, petroleum products, precious metals, and grains). Below is a discussion of exemplary forecasting models and frameworks, including a description of an exemplary engine for consuming news and making asset price forecasts. Performance is examined with the goal of making short-term predictions on returns, number of trades, and volatility.

NMAS can be implemented in a variety of deployments and architectures. For example, within the context of a corporate structure, NMAS data can be delivered via one or more web-based hosted solutions or central servers, or as a deployed solution at a customer or client site through a dedicated service (e.g., an index feed). Figure 1 illustrates an exemplary News/Media Analysis System (NMAS) 100, comprising an online information retrieval system adapted to integrate with either or both a central service provider system or a client-operated processing system. In this exemplary embodiment, NMAS system 100 includes at least one web server that can automatically control one or more aspects of applications on client access devices. It can run applications enhanced with an add-on framework integrated into a graphical user interface or browser control to facilitate interfacing with one or more web-based applications. System 100 includes one or more databases 110, one or more servers 120, and one or more access (e.g., client) devices 130.

News/media database 110 includes a primary (internal) set of databases 112, a secondary (external) set of databases 114, and a metadata module 116. In the exemplary embodiment, internal database 112 includes a news service or database 1121 (represented in this case by the exemplary Thomson Reuters TR News) and a feed service or database(s) 1122 (represented in this case by the exemplary Thomson Reuters TR News Feed). The internal components of news/media database 110 may also include internally sourced social media content. External database 114 includes news (such as and non-internal) services or database(s) 1141, a blog database 1142, a social media database 1143, and other content database(s) 1144. Metadata module 116 includes components adapted to identify, extract, apply, or otherwise recognize metadata associated with news stories and/or social media content. Such metadata may be used by NMAS 100 to pre-process news stories, such as sentence separation, part-of-speech tagging, text parsing, tokenization, etc., to facilitate associating stories with one or more companies and preparing the content for application of computational linguistics processes and sentiment analysis.

The database 110, which takes the exemplary form of one or more electronic, magnetic, or optical data storage devices, includes or is otherwise associated with corresponding indexes (not shown). Each index includes terms and phrases associated with corresponding document addresses, identifiers, and other general information. The database 110 is coupled or couplable to a server 120 via a wireless or wired communication network (such as a local area network, a wide area network, a private network, or a virtual private network).

Server 120, generally representing one or more servers providing data in the form of web pages or other markup languages (with associated applets, ActiveX controls, remote call objects, or other related software and data structures), is configured to serve various "thicknesses" of service clients. More specifically, server 120 includes a processor module 121, a memory module 122, which includes a subscriber database 123, green score/composite index modules 124-125, a user interface module 126, a training/learning module 127, and a classifier module 128. Processor module 121 comprises one or more local or distributed processors, controllers, or virtual machines. Memory module 122, exemplarily in the form of one or more electronic, magnetic, or optical data storage devices, stores subscriber database 123, green score/composite index module 124 (such as for predictive analysis related to a company based on the predictive modeling of the present invention), sentiment processing module 125 (such as for other financial services that users can use to further research companies of interest), and user interface module 126.

Subscriber database 123 includes data related to subscribers used to control, manage, and administer pay-as-you-go or subscription-based access to database 110. In the exemplary embodiment, subscriber database 123 includes one or more user preference (or, more generally, user) data structures 1231, including user identification data 1231A, user subscription data 1231B, and user preferences 1231C. It may also include user-stored data 1231E. In the exemplary embodiment, one or more aspects of the user data structures relate to user customization of various search and interface options. For example, user ID 1231A may include user login and screen name information associated with a user who subscribes to the Green Rating and/or Environmental Composite Index services distributed via NMAS 100. The green score/composite index module 124 includes software and functionality for handling the functions described above and can be applied, for example, in conjunction with one or more of the sentiment processing module 126, the training module 127, and the classifier module 128 against one or more databases 110 to generate or update green scores for companies, or to generate or update composite indices composed of sets of stocks, based on data received from the databases or corpora 110. For example, a training data set or initial data set from the databases 110 applied with some form of validation can be used to train or validate the performance of the NMAS 100 for use in an ongoing manner, such as for use with fee-based services provided by an FSP.

The Information Integration Tool (IIT) framework or interface module 126 (or software framework or platform) includes a set of machine-readable and/or executable instructions for fully or partially defining software and associated user interfaces that integrate or collaborate with one or more applications, including one or more of its components. As shown in FIG2 , the NMAS includes a News/Social Media Processing Engine (NSMPE) that collaborates with IIT 126 and metadata module 116. The NSMPE includes or can collaborate with one or more search engines to receive and process metadata, aggregate, score, filter, recommend, and present results. In an exemplary embodiment, the NSMPE includes one or more feature engines 206, a predictive modeling module 207, a learning or training engine or module 208, and a green scoring and composite indexing module 209 to implement the functionality described herein.

Referring to Figure 1, access device 130 (such as a client device) generally represents one or more access devices. In an exemplary embodiment, access device 130 takes the form of a personal computer, workstation, personal digital assistant, mobile phone, or any other device capable of providing an effective user interface with a server or database. Specifically, access device 130 includes a processor module 131 (one or more processors (or processing circuits) 132), memory 132, a display 133, a keyboard 134, and a graphical pointer or selector 135. Processor module 131 includes one or more processors, processing circuits, or controllers. In an exemplary embodiment, processor module 131 takes any convenient or desired form. Coupled to processor module 131 is memory 132. Memory 132 stores code (machine-readable or executable instructions) for operating system 136, browser 137, and document processing software 138. In an exemplary embodiment, operating system 136 takes the form of a version of the Microsoft Windows operating system, and browser 137 takes the form of a version of Microsoft Internet Explorer. Operating system 136 and browser 137 not only receive input from keyboard 134 and selector 135, but also support rendering of a graphical user interface on display 133. When the process software is launched, an integrated information retrieval graphical user interface 139 is defined in memory 132 and rendered on display 133. When rendered, interface 139 presents data associated with one or more interactive control features (or user interface elements).

In one embodiment using the operating system of the present invention, an add-on framework is installed and one or more tools or APIs on server 120 are loaded onto one or more client devices 130. In an exemplary embodiment, this requires a user to direct a browser on a client access device (such as access device 130) to the Internet Protocol (IP) address for an online information retrieval system (such as offerings from Thomson Reuters Financial and other systems) and then log in to the system using a username and/or password. A successful login results in a web-based interface being output from server 120, stored in memory 132, and displayed by client access device 130. The interface includes an option to initiate the download of information integration software using corresponding toolbar plug-ins for one or more applications. If the download option is initiated, download management software ensures that the client access device is compatible with the information integration software and detects which document processing applications on the access device are compatible with the information integration software. Upon user approval, the appropriate software is downloaded and installed on the client device. In an alternative, an intermediate "enterprise" network server may receive one or more of the framework, tools, APIs, and add-on software for loading onto one or more client devices 130 using internal processes.

Once installed in any manner, the document processing application can then be utilized to present the online tool interface to the user in context. Add-on software for one or more applications can be called simultaneously. The add-on menu includes a list of web services or applications and/or locally hosted tools or services. The user makes a selection via the tool interface, such as manually via a pointing device. Once a selection is made, the selected tool, or more precisely its associated instructions, is executed. In an exemplary embodiment, this requires communication with a corresponding instruction or web application on server 120, which in turn can use one or more APIs stored on the hosted application as part of the add-on framework to provide dynamic scripting and control of the hosted word processing application.

FIG2 illustrates another representation of an exemplary NMAS system 200 for performing the processes described herein, which are executed using a combination of hardware and software, as well as communication networking. In this example, NMAS 200 provides a framework for searching, retrieving, analyzing, and ranking. NMAS 200 can be used in conjunction with a system 204 that provides information or specialized financial services providers (FSPs), such as Thomson Reuters Financial, and includes the information integration and tool framework and application modules 126 described above. Furthermore, in this example, system 200 includes a central network server/database facility 201, which includes a network server 202, a database 203 of documents and information from internal and/or external sources (e.g., news articles, blogs, social media, etc.), an information/document retrieval system 205 (including, as components, a feature building module 206, a predictive module 207, and a training or learning module 208), and a news/social media processing engine including a green scoring and composite indexing engine 209. Central facility 201 can be accessed by remote users 210, such as via a network 226 (e.g., the Internet). Aspects of system 200 can be implemented using any combination of Internet or (World Wide) Web-based, desktop-based, or application Web-enabled components. The remote user system 210 in this example includes a GUI interface operated via a computer 211 (such as a PC computer, etc.), which may include a typical combination of hardware and software, including, as shown with respect to computer 211, system memory 212, operating system 214, application programs 216, graphical user interface (GUI) 218, processor 220, and storage 222. The storage 222 may contain electronic information 224, such as electronic documents and information, such as green score data streams and/or reports, company- and/or industry-based environmental composite index data streams, and/or related reports and information. The methods and systems of the present invention, described in detail below, can be used to provide remote users (such as investors) with access to a searchable database. In particular, remote users can search the database using search queries based on company RICs, green certification lists (as described elsewhere herein), stocks, or other names to retrieve and view predictive analytics and/or recommended actions, as discussed below. RICs refer to Reuters Certification Codes, a label-like code used to identify financial certificates and indexes, and are used to find information on various financial information networks, such as Thomson Reuters market data platforms such as Bridge, Triarch, TIB, and RMDS—the Reuters Market Data System (RMDS) open data integration platform. Green certification lists can take the form of "Green RICs," among others. Client-side application software can be stored on machine-readable media and include, for example, instructions executed by processor 220 of computer 211, and presents web-based interface screens to facilitate interaction between user system 210 and central system 211, such as tools for further analysis of data streams and other data and reports received via network 226 and stored locally or accessed remotely. The operating system 214 should be suitable for use with the system 201 and browser functionality described herein, such as Microsoft Windows Vista (Business, Enterprise, and Ultimate editions), Windows 7, or Windows XP Professional with appropriate service packs. The system may require that the remote user or client machine be compatible with a minimum threshold level of processing power, such as an Intel Pentium III, speed (e.g., 500 MHz), minimum memory level, and other parameters.

The configurations described are, therefore, some of many and do not limit the present invention. Central system 201 may include a network of servers, computers, and databases, such as via a LAN, WLAN, Ethernet, token ring, FDDI ring, or other communication network infrastructure. Any of a number of suitable communication links may be used, such as one or a combination of wireless, LAN, WLAN, ISDN, X.25, DSL, and ATM-type networks. The software used to perform the functions associated with system 201 may include self-contained applications within a desktop, server, or network environment, and may utilize a local database (such as SQL 2005 or above, SQL Express, IBM DB2, or other suitable database) to store documents, collections, and data associated with processing such information. In an exemplary embodiment, the various databases may be relational databases. In the case of relational databases, various data tables are created, and data is inserted into and/or selected from these tables using SQL or some other database query language known in the art. In the case of a database that utilizes tables and SQL, a database application such as MySQL ^™ , SQLServer ^™ , Oracle 81 ^™ , 10G ^™ , or some other suitable database application may be used to manage the data. As is known in the art, these tables may be organized into an RDS or Object-Relational Data Schema (ORDS).

In an exemplary method of the present invention and with reference to the process flow of FIG3 , the following process is performed. First, at step 302 , a user obtains information and content of interest from an appropriate news/social media source (news feed, blog, website, etc.) from an internal or external source. At step 304 , the system applies pre-processing to the obtained information to identify embedded metadata or other descriptors, processing text, words, phrases, and attribute associations about one or more companies. At step 306 , the system applies sentiment analysis and derives one or more sentiment scores associated with the obtained and processed information, as it relates to the companies of interest identified therein. At step 308 , the system may optionally (as discussed elsewhere herein) apply a risk classification method to derive individual scores or indicators associated with a green score or composite index, or to derive a score or indicator. At step 310 , the system applies a predictive model that uses the sentiment scores to derive the green score, for example, to derive a predicted condition or price behavior associated with each company. At step 312, for a set of companies that all have green scores, the system generates an expression of a composite index of the set of green scores, e.g., the index represents the predicted behavior of the corresponding set of stock prices and/or a recommended action to be taken based on the predicted behavior (e.g., buy, sell, or hold).

Figure 4 is a flow chart illustrating database and document processing, sentiment, and green scoring, which utilizes the predictive modeling aspects of the present invention as inputs and outputs for a system employing the present invention, such as the method of Figure 3. For example, external documents, news, social media, and other information (such as news articles and traditional and new media sources, blogs, and social media) are considered inputs to a news/social media processing engine such as the one described above, which may include a combined or separate external messaging engine and an internal data feed messaging engine. Internal news feeds, such as TR Feeds, Reuters News, Westlaw, and Curated feeds, are processed by an internal data feed document processing module. The combined news feeds are further processed by the sentiment scoring engine and ultimately processed according to a predictive model to output a green score for a company and/or a composite index related to the environmental performance or certifications of a set of companies. In this manner, the present invention provides predictive analysis of the corresponding company or other outputs such as a recommended action (buy, sell, or hold). Another output may take the form of a data stream or feed related to the green score or composite index, which may be delivered to subscribers of a financial service and further processed locally. Yet another output may be an intelligent alert service. Additionally, desktop add-ons may include ways to display various outputs and/or receive input in response thereto.

Information-based companies have made significant efforts to collect and/or analyze large corpora or collections of documents and information, including traditional and new-age media, blogs, web pages, and the like. For example, web crawlers and screenshots have been used to extract usable information and data for subsequent processing and analysis, such as formatting/reformatting, structured/unstructured data. Companies can use this information to create or improve the image or identity of their business or products in the minds of customers, which is increasingly important in the context of consumer relations and environmental responsibility. Systems that can identify any underlying "emotions" or "opinions" expressed by expressions in information (e.g., text) are very useful in forming predictive models. This is often referred to as sentiment or opinion mining, and is also known as "feeling" or "emotional" computing. These techniques often utilize natural language processing and are designed to identify and interpret human emotions (opinions, feelings, or emotions, such as happiness, sadness, fear, importance, indifference, positive, negative) and generate responses based on the detected human emotions or feelings.

More specifically, semantic analysis interprets text to identify expressions of sentiment or opinion and can be used to generate semantically aware results. Such systems can be based on ontologies (e.g., the Human Feeling Ontology) and linguistic resources (e.g., WordNet-Affect (WNA)). By extending the use of the system beyond traditional news sources, NMAS can employ the technology to interpret and process opinions and emotions expressed in non-traditional channels/sources (such as blogs, wikis, online forums, message boards, chat rooms, social media networks, etc.) in order to determine green sentiment and green scores. Utilizing all media sources, and particularly for "new media" sources that lack historical internal verification processes, the system can also assign a certain level of verification regarding the (actual or perceived (short-term)) accuracy of the message. Furthermore, the system can be configured to identify "fake" news and anticipate the short-term effects of such "news" when predicting stock price behavior.

By way of example, the sentiment scoring functionality described herein can be performed by the Reuters NewsScope Sentiment Engine (RNSE). RNSE enables clients to leverage a unique set of news/social media sentiment, correlation, and novelty indicators for algorithmic trading systems, as well as risk management and human decision support processes. The service utilizes a linguistic model that scores sentiment in milliseconds for news/social media feeds related to 40 commodity and energy assets and over 10,000 companies currently supported in the offering. Algorithmic trading is useful for both buy-side and sell-side market participants in cash equity markets, as well as other liquid asset classes such as foreign exchange, commodities, and energy. Commodity markets offer numerous opportunities for institutional investors and proprietary traders to grow and diversify their investment strategies. Given the growth, price volatility, and increasing adoption of this asset class in active trading strategies across global commodity and energy markets, client demand for relevant quantitative solutions continues to grow. The sentiment scores and resulting green scores or composite indices can be used by trading desks and quantitative research analysts to better model asset price movements. Clients have access to historical data, allowing them to backtest their systems for suitability against their trading and investment strategies.

FIG5 is a flowchart illustrating the steps in an exemplary method for generating sentiment for use in green scoring, such as for benchmarking the greenness of public and private companies using social media and news content. Exemplary data sources for processing by NMAS 100 include: news agency wire sources (e.g., AFP, AP, TR, Reuters, Bloomberg), social media (blogs, Twitter, RSS, Gigaom, NWCleanTech, ClimateWire), and internet/web-based sources (e.g., CNN.com, WSJ.com, lesoir.be). In today's environment, social media often provides a more timely source of information than traditional news channels. For example, a blogger may post a comment about "Company A," and the comment and further comments may be noticed on a social media feed before the company is ultimately mentioned in syndication and traditional news reports/sources. This is particularly true in the case of "green" issues and content. By examining social media-based sentiment, the present invention can be more responsive in predicting corporate behavior and stock prices regarding green issues. In the example of FIG5 , the following analyses are performed: entity extraction (e.g., object, company, location, etc.), source, author, number of news items, relevance to a particular taxonomy/theme (e.g., green), fact extraction, topic code assignment, classification assignment, analytical tone, assigned sentiment (+ or -), certification code assignment (e.g., RIC, green RIC). The output resulting from analyzing the source data can be delivered in any of the following forms: a real-time stream (and historical database) of sentiment/scores for a given company for a given taxonomy; a real-time stream (and historical database) of sentiment/scores for more than one company in a composite composite index; an alert service in the form of an electronic message indicating that the index for a certain company has significantly more than a preset % within a given time period; and/or an alert service in the form of an electronic message indicating that the index for a certain company has significantly more than a preset % within a given time period preset by the user/system. The recipient of the deliverable output can then further process the output as desired.

Figure 6 is a diagram illustrating a representation of a green community in the form of a website. The community can include access to and leverage existing resources and tools. For example, the community includes aggregated assets, analytics and tool assets, and distributed assets to provide a robust and efficient experience for users (such as investors and those in the investment community). In this example, aggregated assets include: News; StarMine; Legal Entities; GRID; NOVUS; Social Media; Website; Crowdsourcing Software; Moreover/InfoEngine. Analytics assets may include: News Sentiment Engine; OpenCalais; Lipper Benchmarks; Velocity Analytics; Machine Learning Tools; Green Sentiment; Green Taxonomy; Extensive Text Analytics (Lexalytics); and Alerts (Psydex). Distributed assets may include: Eikon/Omaha; DataScope; Elektron; Enterprise Services Portal; Content Marketplace; IDN/RIC/RFA; Reuters.com Blog; News Archive; (One or More) Green Websites; and Blog Community.

Using the NMAS 100 system and related technologies described herein, the present invention addresses a broad set of needs by providing intelligent information and analytical tools to monitor and predict the impact of green behavior at the company and index level. The present invention can be used to access a historical database of green news tagged to individual companies, track real-time alerts for major news with associated green scores, monitor social media feeds and track green initiatives or events, publish/receive green sentiment scores for different companies, and utilize group tools to monitor peer behavior. Green asset managers can use the present invention to implement and monitor adherence to green investment goals and requirements and identify alpha generation strategies. Enterprises can use the present invention in a more inward-directed manner for brand monitoring and for implementing and evaluating CSR and other related initiatives. Regulatory agencies (e.g., the Environmental Protection Agency) can use the present invention to monitor and supervise green compliance and for input into green legislation.

Referring now to FIG. 7 , and in relation to the present invention's Green Sentiment Composite Index, the NMAS 100, as its core foundation, may incorporate a combination of machine learning and artificial intelligence (AI) capabilities, providing intelligent information for use in analyzing the impact of green behavior by public and private companies. The resulting outputs of the NMAS 100 may take the form of green sentiment company and composite indexes, intelligent alerts, and/or a desktop client/interface and toolset. The NMAS 100 may utilize highly specialized taxonomies specifically for scoring environmental attributes relevant to companies and industries. Each source will have its own nuanced taxonomy and weighting for index calculation (e.g., performed by Velocity Analytics). Once operational, the AI can adapt to changing market conditions, expanding the taxonomy to include newly developed lingo and highlighting textual patterns most relevant to equity price movements. In implementation, the present invention can provide a categorization for green investments, trigger green alerts within the SEC, enable investors to trade based on green RICs or categorizations, add social media components to the overall green investment community, and deliver green data feeds for further processing by investors.

Services like InfoEngine offer out-of-the-box aggregation of Twitter, blogs, online news feeds, and other types of third-party content. For example, content aggregators like InfoEngine, computational engines like Lexalytics, and community websites are used. Once the feeds are fed into the server, OpenCalais/ClearForest, for example, is used for smart tagging, which helps distinguish between feeds. Once the taxonomy and corresponding algorithms are applied, computational engines like Lexalytics will then score the articles.

Sentiment scores from different sources will be weighted based on their importance. Widely circulated online and newswire sources will be weighted based on their Alexa and Nielsen ratings, while social media sources will be weighted based on their followers, subscribers, and impressions. The weighted scores will then be aggregated to provide an overall "green sentiment." Similar to the evolution of the taxonomy, weights can be changed as AI detects a higher correlation between a source and a company's equity price. Finally, a community website will be built to promote the green social media debate and will be used to maintain the green taxonomy.

Risk Mining

8-16 are examples of risk mining techniques for implementing the present invention. Risk mining techniques will be described more fully below for use in conjunction with the present invention.

Figure 8 illustrates how risks materialize over time. Initially, a risk P=>Q is extracted from a large text database, where Q represents a high-impact event and P represents a prerequisite for Q, which is causally or statistically connected to Q and precedes Q in time. Unless otherwise stated or indicated herein, the implication symbol "=>" captures the causal and/or enabling relationship that exists between P and Q (e.g., P causes Q, or P may enable Q). The implication symbol "=>" does not mean substantive implication. Later, at time t.sub.j, P may occur, which in turn may cause Q to occur at time t.sub.k. The present invention solves the problem of automatically obtaining the risk P=>Q from text, and describes how P=>Q and P can be used to alert a user that Q may be imminent. As used herein, the term "risk", which can be positive or negative, refers to an event involving uncertainty (unless the event has already occurred) that may be caused by a factor, thing, element or process. In particular, as used herein, the term "risk," which may be positive or negative, refers to a condition that is a priori for an event, wherein the condition is causally or statistically linked to the event and temporally precedes the event. As used herein, the term "precondition" refers to a statement or indication related to a particular object. In particular, the term "precondition" refers to a statement or indication that is related to a particular event, either directly or through the mining techniques of the present invention.

A corpus (e.g., a collection of one or more text feeds) is mined for risk using a computing device. As used herein, the term "corpus" and variations thereof refer to one or more data sets, particularly digital data including text data. The corpus may include, but is not limited to: news; financial information, including but not limited to stock price data and its standard deviation (instability); government and regulatory reports, including but not limited to government agency reports, regulatory filings such as tax filings, medical filings, legal filings, Food and Drug Administration (FDA) filings, and Securities and Exchange Commission (SEC) filings; private entity publications, including but not limited to annual reports, newsletters, advertisements, and press releases; blogs; web pages; event streams; agreement documents; status updates on social networking services; emails; short message service (SMS); instant chat messages; Twitter tweets; and/or combinations thereof. The computing device investigates the corpus to extract risk-indicating patterns and utilizes the risk-indicating seed patterns as seeds for a risk identification algorithm for subsequent risk mining by an analyst or user. The computing device may also include an interface (such as a keyboard) for querying the computer and a display for displaying results from the computer.

The computing device can also be used to alert a user to risks, including but not limited to imminent risks, i.e., risks that are likely to occur, including but not limited to those that are likely to occur in the near future or within a defined time period, via a computer interface (not shown). The user is typically alerted via a computing device (not shown). However, the present invention is not limited to this, and any device with a visual display or even voice communication may be suitably used. As used herein, the term "computing device" refers to a device that performs calculations, particularly a programmable electronic machine that performs high-speed mathematical or logical operations or aggregates, stores, correlates or otherwise processes information. Examples include (without limitation) mainframe computers, personal computers and handheld devices. Prior to mining a corpus for risks, the present invention utilizes a computing device to extract risk indication patterns from one or more corpora of text data. As used herein, a risk indication pattern is a pattern developed through the techniques of the present invention that relates possible prerequisites to possible events.

The computing device includes a risk identification algorithm. Using the computing device including the risk identification algorithm, a corpus of textual data is searched by a risk miner for instances of a set of risk-indicating seed patterns provided to create a risk database. The corpus may include, but is not limited to: news; financial information, including but not limited to stock price data and its standard deviation (instability); government and regulatory reports, including but not limited to government agency reports, regulatory filings such as tax filings, medical filings, legal filings, Food and Drug Administration (FDA) filings, and Securities and Exchange Commission (SEC) filings; private entity publications, including but not limited to annual reports, newsletters, advertisements, and press releases; blogs; web pages; event streams; protocol documents; status updates on social networking services; emails; short message services (SMS); instant chat messages; Twitter tweets; and/or combinations thereof. Corpus 210 may be the same as or different from corpus 110.

In one embodiment of the present invention, a risk database is generated using trigger keywords (e.g., "risk," "threat"). In another embodiment, a risk database is generated using regular expressions (e.g., "("may")?pose(s)?(a)?threat(s)?to"). Candidate risk statements or statement sequences are created, and new patterns are generalized by running a named entity tagger or part-of-speech (POS) tagger and chunker on them (entities can be described by proper nouns or NPs, not just given by named entities), and replacing entities with placeholders for each category (e.g., "J.P. Morgan" => "<COMPANY>"). These generated patterns can be used to reprocess the corpus, either after some human review in one embodiment of the present invention, or automatically in another embodiment. The extracted statement or sequence of statements is then both verified (whether it is really a risk-indicating statement) and parsed into a risk of the form P=>Q (i.e., finding out which text spans correspond to the premise "P", which parts express the entailment "=>", and which parts express the high-impact event "Q"), using but not limited to the following non-limiting features: a set of terms that have significant statistical association with the term "risk" (in one embodiment of the invention, statistical procedures such as pointwise mutual information (PMI) and log-likelihood or rules including but not limited to rules obtained by Hearst pattern induction are used to determine the term set); a binary gazetteer feature set, where the feature is excited if the gazetteer is compiled by a human expert or a set of risk-indicating terms ("threat", "bankruptcy", "risk"...) extracted from manually labeled training data; a set of indicators of speculative language; instances of future time references; the occurrence of conditions; and/or the occurrence of causal markers.

In one embodiment of the present invention, a variation of surrogate machine learning (i.e., a technique for learning a task by learning from examples) can be used to create training data for a machine learning-based classifier for extracting risk-indicator statements. A useful technique is described by Sriharsha Veeramachaneni and Ravi Kumar Kondadadi in "Surrogate Learning—From Feature Independence to Semi-Supervised Classification," Proceedings of the NAACL HLT Workshop on Semi-supervised Learning for Natural Language Processing, pp. 10-18, Boulder, Colo., June 2009, Association for Computational Linguistics (ACL), the contents of which are incorporated herein by reference.

The risk type classifier classifies each risk pattern by risk type ("RT") according to a predefined taxonomy of risk types. In one embodiment of the invention, the taxonomy may use, but is not limited to, the following non-limiting categories: Political: government policy, public opinion, ideological change, doctrine, legislation, unrest (war, terrorism, riots); Environmental: contaminated land or pollution liability, nuisance (e.g., noise), permits, public opinion, internal/corporate policy, environmental laws or regulations or practices or "impact" requirements; Planning: permit requirements, policies and practices, land use, socioeconomic impacts, public opinion; Market: demand (forecast), competition, obsolescence, customer satisfaction, fashion; Economic: fiscal policy, taxes, cost inflation, interest rates, exchange rates; Finance: bankruptcy, profits, insurance, Risk sharing; Physical: unforeseen ground conditions, weather, earthquakes, fire, explosions, archaeological discoveries; Project: definition, procurement strategy, performance requirements, standards, leadership, organization (maturity, commitment, competence and experience), planning and quality control, procedures, workforce and resources, communication and culture; Technical: design completeness, operational efficiency, reliability; Regulatory: changes by governing bodies; Human: error, incompetence, neglect, fatigue, communication skills, culture, working in the dark or at night; Criminal: lack of security, vandalism, theft, fraud, corruption; Safety: regulations, hazardous materials, collisions, collapse, floods, fire, explosions; and/or Legal: changes in legislation, treaties.

The risk clusterer groups all risks in the database by similarity, without imposing a predefined taxonomy (data-driven). In one embodiment, Hearst Schema Induction can be used. Hearst Schema Induction was first described in Hearst, Marti, "WordNet: An Electronic Lexical Database and Some of its Applications" (Christiane Fellbaum, MIT Press, 1998), the contents of which are incorporated herein by reference. In another embodiment of the present invention, the number k is chosen by the system developer, and a kNN means clustering method can be used. Further details of kNN clustering are described in Hastie, Trevor, Robert Tibshirani, and Jerome Friedman, "The Elements of Statistical Learning: Data Minig, Inference, and Prediction" (2nd edition, Springer, 2009), the contents of which are incorporated herein by reference. In such a case, risks are grouped into a certain number (i.e., k) of categories and then classified by selecting the cluster with the highest similarity to the cluster of interest. In another embodiment of the present invention, hierarchical clustering is used. Alternatively or additionally, both k-means clustering and hierarchical clustering may be used.

In one embodiment of a risk clusterer according to the present invention, a corpus of text is provided. The corpus of text is tokenized into a set of sentences. All instances of risks indicated by "*" are extracted from the tokenized text. A taxonomy of risks is constructed into a tree by organizing all filler characters (i.e., "*") that match the risk. Hearst pattern induction can be used to summarize the risk taxonomy. In addition, an NP chunker can be used to find interesting boundaries.

In another embodiment of a risk clusterer according to the present invention, a risk taxonomy is created from, for example, risks, legal risks and legal changes. As indicated by , risks such as those that may be associated with legal changes are taken as seeds. As indicated by , legal risks such as legal changes are mined by a computing device. As indicated by , risks are also mined for legal risks. In such a manner, based on risks and legal changes, there is feedback for legal risks. Mining for risks and legal risks may include mining using the word risk or its equivalents. Mining for legal changes does not necessarily include the word risk. Advantageously, the taxonomy resulting from this process contains risk-indicative phrases that do not necessarily contain the word "risk" itself. In addition to its use for risk type classification, such a taxonomy may also be used in a risk mining model.

The risk alerter performs a similarity matching operation between the risks in the database and possible instances of P or Q in the text feed 110. If evidence is found for P, the risk P=>Q is "imminent." If evidence is found for Q, the risk P=>Q has materialized. In one embodiment of the present invention, the risk alerter directly delivers a warning notification to the user.

Thus, when examining a risk database, a user (e.g., a risk analyst) can take immediate action before risks materialize and prioritize the management of both the impending risks in the text feed ("P!,...,P!,P!,P!,...P!...") and the materialized risks ("Q!") as events unfold, without even having to read the text feed.

In one embodiment of the present invention, the output of a risk alerter is connected to the input of a risk routing unit. The risk alerter notifies an analyst that their profile matches a risk type RT. For example, an analyst may wish to learn about environmental risks. When the preconditions for a possible environmental event are discovered, the risk alerter will alert the analyst to the environmental risk. For example, if industrial activity increases in a particular country or region, the analyst may be redirected to the environmental risk of global warming.

In one embodiment of the present invention, a set of risk profiles, such as those extracted from a complete set defined as a set of all past Securities and Exchange Commission ("SEC") filings, is matched to risks extracted from a text feed. To ensure compliance with SEC business risk disclosure obligations, the method proposes a ranked list of risk profiles or alternative risk profiles for inclusion in a draft SEC filing for a company operating the system.

The present invention may use a variety of methods for risk identification. For example, as depicted in Figure 9, risk mining may include: baseline monitoring of regular patterns on surface strings and named entity labels; identification of words frequently associated with risk using cluster information theory; and/or clustering of risk-indicating statements. Alternatively or in addition, techniques for machine learning of tasks using examples may be used. Risk identification includes querying one or more full sets of risk-indicating patterns. The query results may match all, substantially all, or some of the risk-indicating patterns. Occurrence counts or specific risk-indicating patterns may also be used in the risk mining techniques of the present invention.

10 and 11 illustrate examples of risk mining according to the present invention. In Example 1 of FIG. 10 , the full set of listed news articles is mined for the term “cholesterol” which is a prerequisite or P for a Q or event. The event Q is further classified by the holder “diabetics” and the target “amputation risk”. The risk type RT is health and has a positive polarity because it is beneficial to health. For the purposes of the present invention, the term “risk” refers not only to negative or harmful events, but also to positive or beneficial outcomes. In other words, a risk can have positive and/or negative effects. In Example 2 of FIG. 11 , the full set of listed news articles is mined for the term “North Korea launch” which is a prerequisite or P for a Q or event. Event Q is further categorized by the subject "North Korea" and the target "more than condemnation" U.S. The risk type RT is political and has a negative polarity due to its harmfulness to world politics. Furthermore, such negative and/or positive polarities can be weighted according to the degree of risk. In such cases, it may be beneficial to change the risk of user 130 to a greater degree, such as being very harmful or very beneficial, for a risk with lesser consequences.

FIG12 illustrates another example of risk mining according to the present invention. In Example 3, a news article is mined. As background, the demand for metallic lithium increases when a limited supply is available. Much of the metal is obtained from Bolivia, a country whose government may be considered by some to be unfriendly to capitalist governments or companies at the time the article was published. As indicated by the underlined words and/or sequences, the article is mined for various potential words, word sequences, and/or partial phrases to query the article for prerequisites P of an event Q that may lead to a risk. The types of risks present in the article include supply and demand risks and political risks.

FIG13 illustrates another example of risk mining according to the present invention. In Example 4a, a full set of patterns with specific flags, namely "if" and "then," is mined. Mining extracts sequences that begin or have these flags. The length of the sequence is not limited to any specific length or number of words, but is determined by the flags. The sequence is stored, for example, in a register within a computing device. However, the use of patterns such as, but not limited to, those shown in FIG16 can be more accurate than using keyword-based ranking searches.

FIG14 illustrates another example of risk mining according to the present invention. In Example 5a, the corpus is mined based on the syntactic or grammatical structure of sentences or phrases. In this example, the standard PENN Treebank classification or labeling or slightly modified PENN labels are used. Further details of the Penn Treebank can be found at http://www.cis.upenn.edu/.about.treebank/ (the PENN Treebank homepage), the contents of which are incorporated herein by reference, or by contacting the Linguistic Data Consortium, University of Pennsylvania, 3600 Market Street, Suite 810, Philadelphia, Pa. 18104. Corresponding label sets have been established for languages other than English and are known to those skilled in the art. In this example, the label "PRP" refers to personal pronouns, i.e., "we" in the example sentence. The label "VBP" refers to non-third person singular present tense verbs, i.e., "expect" in the example sentence. The label "TO" simply refers to the word "to" in the example sentence. The "VB" tag refers to the infinitive verb, which is "be" in the example sentence. The "RB" tag refers to the adverb, which is "negatively" in the example sentence. The "IN" tag refers to the preposition or subordinating conjunction, which is "by" in the example sentence. Some common PENN Treebank word P.O.S. tags include but are not limited to: CC - coordinating conjunction; CD - cardinal number; DT - determiner; EX - existential; FW - foreign word; IN - preposition or subordinating conjunction; JJ - adjective; JJR - comparative adjective; JJS - superlative adjective; LS - list item marker; MD - modal verb; NN - noun, singular or uncountable; NNS - plural noun; NNP - proper noun singular; NNPS - proper noun plural; PDT - prepositive determiner; POS - possessive ending; PRP - personal pronoun; PRP$ - possessive pronouns (prelog version PRP-S); RB—adverb; RBR—comparative adverb; RBS—superlative adverb; RP—particle; SYM—sign; TO—to; UH—interjection; VB—infinitive; VBD—past tense verb; VBG—verb, gerund, or present participle; VBN—past participle verb; VBP—verb, non-third person singular present tense; VBZ—verb, third person singular present tense; WDT—Wh determiner; WP—Wh pronoun; WP$—possessive wh pronoun (prelog version WP-S); and WRB—Wh adverb.

In Figure 15, Example 6 illustrates another mining sequence or algorithm based on the PENN treebank tags. Thus, as shown in Figures 14 and 15, the mining technology of the present invention can analyze the same statement under different criteria to obtain risks or prerequisites for risks.

In FIG16 , risk mining according to the present invention is performed through a sequence of binary grammatical dependency relationships between words (including placeholders).

The examples and techniques described above for mining risks can be used individually or in any combination. However, the present invention is not limited to these specific examples, and other patterns or techniques can be used with the present invention. The mined patterns from these examples and/or from the techniques of the present invention can be ranked according to various ranking algorithms, such as, but not limited to, statistical language models (LMs), graph-based algorithms (such as PageRank or HITS), ranked SVMs, or other suitable methods.

In one aspect of the present invention, a computer-implemented method for mining risks is provided. The method includes providing a set of risk indication patterns on a computing device; querying the full set using the computing device to identify a set of potential risks using a risk identification algorithm based at least in part on the set of risk indication patterns associated with the full set; comparing the potential risk set with the risk indication patterns to obtain a set of prerequisite risks; generating a signal representing the set of prerequisite risks; and storing the signal representing the set of prerequisite risks in an electronic memory. The method may also include determining an impending risk based on the prerequisite risks, the impending risk determined using the risk identification algorithm, the impending risk associated with at least one risk from the set of prerequisite risks; generating a signal representing the impending risk; and storing the signal representing the impending risk in the electronic memory. Furthermore, the method may also include determining a materialized risk after storing the signal representing the set of prerequisite risks, the materialized risk determined using the risk identification algorithm, the materialized risk associated with the set of risks; generating a signal representing the materialized risk; and storing the signal representing the materialized risk in the electronic memory. In addition, the method may further include: determining a materialized risk after storing the signal representing the imminent risk, the materialized risk being determined using the risk identification algorithm, the materialized risk being associated with the imminent risk; generating a signal representing the materialized risk; and storing the signal representing the materialized risk in the electronic memory.

Desirably, the corpus is digital. The corpus may include, but is not limited to: news; financial information, including but not limited to stock price data and its standard deviation (instability); government and regulatory reports, including but not limited to government agency reports, regulatory filings such as tax filings, medical filings, legal filings, Food and Drug Administration (FDA) filings, Securities and Exchange Commission (SEC) filings, and the like; private entity publications, including but not limited to annual reports, newsletters, advertisements, and press releases; blogs; web pages; event streams; protocol documents; status updates on social networking services; emails; short message services (SMS); instant chat messages; Twitter tweets; and/or combinations thereof.

The risk identification algorithm can be based on various factors and/or criteria. For example, the risk identification algorithm can be based on, but is not limited to: a set of terms statistically associated with risk; a time factor; a customized set of criteria; and combinations thereof. The customized set of criteria can, for example, include and/or consider: industry criteria, geographic criteria, monetary criteria, political criteria, severity criteria, urgency criteria, subject matter criteria, topic criteria, a named entity set, and combinations thereof.

In one aspect of the present invention, the risk identification algorithm can be based on a source rating set. As used herein, the phrase "source rating" refers to a rating of a source, such as, but not limited to, relevance, reliability, and the like. A source rating set can have a one-to-one correspondence with a source set. A source set can serve as the source of information upon which the full set is based. The source rating set can be modified based on impending risk, materialized risk, or a combination thereof.

The method of the present invention may further include transmitting a signal representing the set of prerequisite risks, transmitting a signal representing the impending risk, transmitting a signal representing the materialized risk, and combinations thereof. Furthermore, the present invention may further include providing a web-based risk alert service using at least one of the following: a signal representing the set of risks, a signal representing the impending risk, a signal representing the materialized risk, and combinations thereof.

In another aspect of the present invention, a computing device may include: electronic storage; and a risk identification algorithm based at least in part on a set of risk indicative patterns associated with a full set stored in the electronic storage. A processor (not shown) may be used to execute the algorithm on the computing device. The computing device may include a computer interface, depicted as (but not limited to) a keyboard, for querying the risk identification algorithm. The computing device may also include a display for receiving signals from the electronic storage and for displaying risk alerts from the risk identification algorithm.

In another aspect of the present invention, a computer system for alerting a user to a risk is provided. The system may include a computing device having electronic storage and a risk identification algorithm, the risk identification algorithm being based at least in part on a set of risk-indicating patterns associated with a full set stored in the electronic storage. The algorithm may be used to execute the algorithm on the computing device. The system may also include a user interface for querying the risk identification algorithm and for receiving a signal from the computing device's electronic storage for alerting the user to the risk. The user interface may include, but is not limited to, a computer, a television, a portable media device, and/or a web-enabled device such as a cellular phone, a personal digital assistant, or the like.

In implementation, the concepts of the present invention may be performed automatically or semi-automatically, i.e., with some degree of human intervention. Likewise, the present invention is not limited in scope by the specific embodiments described herein. It should be fully contemplated that various other embodiments and modifications to the present invention besides those described herein will become apparent to those skilled in the art based on the foregoing description and accompanying drawings. Therefore, such other embodiments and modifications should be intended to fall within the scope of the following claims. Furthermore, although the present invention is described herein in the context of specific embodiments and implementations and applications and in specific environments, those skilled in the art will recognize that its usefulness is not limited thereto, and that the present invention may be beneficially applied in any number of ways and environments for any number of purposes. Therefore, the claims set forth below should be interpreted in accordance with the full scope and spirit of the invention as disclosed herein.

Claims (46)

1. A computer-implemented method, comprising: (a) The entity to which green points will be assigned will be identified by computer; (b) Perform sentiment analysis on a first set of social media information related to the identified entity, the sentiment analysis including text-based sentiment analysis of the first set of social media information; (c) A sentiment score is determined by computer based at least in part on a sentiment analysis performed, including the text-based sentiment analysis described above; (d) A first green score is calculated by a computer based at least in part on a determined emotion score and a predictive model applied to the determined emotion score, wherein the first green score represents a first green emotion score; (e) Transmit the green score via computer; (f) Subsequently, sentiment analysis is performed on a second set of social media information related to the identified entity, the sentiment analysis including text-based sentiment analysis of the first set of social media information; (g) A second sentiment score is determined by computer based at least in part on a sentiment analysis performed, including the text-based sentiment analysis; (h) A second green score is calculated by a computer based at least in part on a determined second emotion score and a predictive model applied to the determined second emotion score, wherein the second green score represents a second green emotion score; and (i) The first green sentiment score and the second green sentiment score are weighted based on their importance, and the weighted sentiment scores are aggregated to provide a green score. 2. The method of claim 1, wherein the social media information set is associated with an identified entity, the entity is associated with traded securities, the social media information set represents green information relating to the entity, and the first or second green score represents the environmental impact of the entity. 3. The method of claim 1, further comprising: modifying the green score over time using a computer based on an additional set of social media information, and transmitting the modified green score. 4. The method of claim 1, wherein calculating the green score by computer comprises: Text is identified by a computer within a set of content received from a social media collection, and the text is identified as representing the entity. A sentiment score is determined by computer based on text-based sentiment analysis, at least in part, based on the identified text; and The green score refers to the green mood score. 5. The method of claim 1, further comprising: certifying the entity as green compliant at least in part based on the green score. 6. The method of claim 1, further comprising: performing steps (a)-(c) via a computer to result in a second green score for the second identified entity, and wherein transmitting the green score includes transmitting a data feed via a computer including data reflecting the green score and the second green score. 7. The method of claim 1, further comprising: generating an alarm signal relating to the entity by means of a computer based at least in part on the calculated value of the green score. 8. The method of claim 1, wherein identifying an entity by computer further includes identifying text by computer in a content set received from a social media set, the text being identified as representing the entity, and further includes extracting additional text by computer from the content set that is identified as representing emotions related to the environmental impact of the entity. 9. The method of claim 8, wherein identifying the text in the content set by computer includes one or more of the following: identifying embedded metadata or other descriptors by computer; processing text, words, or phrases by computer; applying natural language linguistic analysis by computer; and applying Bayesian techniques by computer. 10. The method of claim 1, further comprising aggregating a set of content from a plurality of sources via computer, the plurality of sources including at least one social media source and at least one additional source from the group consisting of: news websites; online forums; government agency websites; academic institution/political party websites; online magazine websites; blog websites; microblogging websites; social and professional networking sites; online publicity and fundraising websites; Facebook; Twitter; and an information aggregator, wherein the information aggregator aggregates content from a set of sources from the plurality of sources, the set of content including the social media information set. 11. The method of claim 1, further comprising: (d) Identify by computer the set of companies to be associated with the composite index, the set of companies including the entities and associated with the set of securities; (e) Generating a composite contextual index for the security set by computer, at least in part based on the social media information set; and (f) Transmitting signals associated with the composite environment index via a computer. 12. The method of claim 11, wherein the composite environment index is determined at least in part based on a set of green scores associated with the set of companies. 13. The method of claim 1, wherein the green score is generated in real time. 14. The method of claim 1, wherein the green score is obtained by computer based on one or more of the following positive criteria: compliance or certification related to the product or manufacturing environment; energy efficiency; corporate practices that promote environmental management, consumer protection, human rights and diversity; business/products involved in green technologies, energy-efficient technologies, alternative fuel technologies, renewable resource technologies; and/or based on one or more of the following negative criteria: business involved in alcohol, tobacco, gambling, weapons and/or military matters; and business that does not comply with environmental standards. 15. The method of claim 1, wherein the entity is associated with securities traded in the market, and further comprises obtaining predictive behavior associated with the securities by applying a predictive model via computer. 16. The method of claim 15, further comprising: generating by computer an expression of the predicted behavior and/or a suggested action to be taken based on the predicted behavior. 17. The method of claim 16, wherein the recommended action relates to a trading decision concerning the security and is one of the group consisting of buying, selling, or holding. 18. The method of claim 1, wherein the social media information set is identified by a computer based on time value. 19. The method of claim 1, further comprising: generating a risk signal representing a potential risk via a computer. 20. The method of claim 1, further comprising: Provide a set of risk indication patterns on computing devices; and A set of potential risks associated with the entity is identified by computer within the social media information set by using a risk identification algorithm that is at least partially based on the risk indication pattern set. 21. The method of claim 20, further comprising: The potential risk set is compared with the risk indication pattern by a computer to obtain the prerequisite risk set; A signal representing the prerequisite risk set is generated by computer; and The signal representing the prerequisite risk set is stored in an electronic memory by a computer. 22. The method of claim 1, further comprising: Repeating steps (a)-(c) by computer on the entity set results in a green score calculated for each entity in the entity set; Green classification is identified by computer; and Entities are selected from the set of entities included in the green category by means of a computer, at least in part, based on the corresponding entity green score. 23. The method of claim 22, wherein the green classification involves certifying companies as green compliant via computer, and wherein each of the selected entities is certified as green compliant. 24. A computer-based system, comprising: A processor adapted to execute code; Memory used to store executable code; Input is adapted to receive a set of social media messages associated with the entity to which green points will be assigned; A sentiment analysis module adapted for execution by a processor, the sentiment analysis module including code executable by the processor to perform sentiment analysis on a set of social media information related to the identified entity, the sentiment analysis including text-based sentiment analysis of the social media information set; A sentiment score module adapted for execution by a processor, the sentiment score module including code executable by the processor to determine a sentiment score based at least in part on the executed sentiment analysis including the text-based sentiment analysis; A green score module adapted for execution by a processor, the green score module including processor-executable code to calculate a first green score based at least in part on a determined sentiment score, wherein the first green score represents a first green sentiment score; and The output is adapted to transmit a signal associated with the green score; and The sentiment analysis module is also adapted to subsequently perform sentiment analysis on a second set of social media information associated with the identified entity, the sentiment analysis including text-based sentiment analysis of the first set of social media information; The emotion score module is also adapted to determine a second emotion score based at least in part on an performed emotion analysis, including the text-based emotion analysis; and The green score module is further adapted to: calculate a second green score based at least in part on a determined second emotion score, wherein the second green score represents a second green emotion score; weight the first green emotion score and the second green emotion score based on their importance; and aggregate the weighted emotion scores to provide a green score. 25. The system of claim 24, wherein the entity is associated with traded securities, the social media information set represents green information relating to the entity, and the first or second green score represents the entity’s environmental impact. 26. The system of claim 24, wherein the green score module includes code executable by a processor to modify the green score over time based on an additional set of social media information, and wherein the output is adapted to transmit the modified green score. 27. The system of claim 24, further comprising an emotion module adapted for execution by a processor, and the emotion module comprising code executable by the processor to: The text in the social media information set is identified as representing the entity; and The sentiment score is determined at least in part based on text-based sentiment analysis of the identified text, wherein the green score represents a green sentiment score. 28. The system of claim 24, further comprising means for certifying the entity as green compliant at least in part based on a green score. 29. The system of claim 24, wherein the output is adapted to transmit a data feed including data reflecting a plurality of green scores. 30. The system of claim 24 further includes code executable by a processor to generate an alarm signal relating to the entity, at least in part based on the calculated green score value. 31. The system of claim 24, further comprising code executable by a processor to identify text in the social media information set as representing the entity, and executable to extract text from the social media information set identified as representing emotions related to the environmental influence of the entity. 32. The method of claim 31 further includes code executable by a processor to: identify the embedded metadata or other descriptors; process text, words, phrases; and apply natural language linguistic analysis. 33. The system of claim 24, further comprising processor-executable code for aggregating a set of content from a plurality of sources, the plurality of sources including at least one social media source and at least one additional source from the group consisting of: news websites; online forums; government agency websites; academic institution/political party websites; online magazine websites; blog websites; microblogging websites; social and professional networking sites; online publicity and fundraising websites; Facebook; Twitter; and information aggregators, wherein the information aggregator aggregates content from a set of sources from the plurality of sources, the set of content including the social media information set. 34. The system of claim 24, further comprising a composite index module adapted to identify a set of companies to be associated with the composite index, the set of companies including the entities and associated with a set of securities, the composite index module further adapted to generate a composite context index for the set of securities based at least in part on the set of social media information, and wherein the output is adapted to transmit a signal associated with the composite context index. 35. The system of claim 34, wherein the composite index module includes code executable by a processor to determine the composite environment index based at least in part on a set of green scores associated with the company set. 36. The system of claim 24, wherein the green score is generated in real time. 37. The system of claim 24, wherein the green score is obtained based on one or more of the following positive criteria: compliance or certification related to the product or manufacturing environment; energy efficiency; corporate practices that promote environmental management, consumer protection, human rights and diversity; business/products involved in green technologies, energy-efficient technologies, alternative fuel technologies, renewable resource technologies; and/or based on one or more of the following negative criteria: business involved in alcohol, tobacco, gambling, weapons and/or military matters; and business that does not comply with environmental standards. 38. The system of claim 24, wherein the entity is associated with a security traded in the market, and further includes code executable by a processor to apply a predictive model or classification method to obtain predictive behavior associated with the security. 39. The system of claim 38, wherein the output is adapted to transmit a signal representing an expression of the predicted behavior and/or a suggested action to be taken based on the predicted behavior. 40. The system of claim 39, wherein the recommended action relates to a trading decision concerning the security and is one of the group consisting of buying, selling, or holding. 41. The system of claim 24, wherein the social media information set is identified based on time value. 42. The system of claim 24 further includes code executable by a processor to generate a risk signal representing a potential risk. 43. The system of claim 24, further comprising: A set of risk indication patterns stored in memory and accessible by the processor; and A risk identification module with code that can be executed by a processor to identify a set of potential risks associated with the entity within the social media information set by using a risk identification algorithm that is at least partially based on the risk indication pattern set. 44. The system of claim 43, wherein the risk identification module includes code executable by a processor to perform the following operations: comparing the potential risk set with the risk indication pattern to obtain a prerequisite risk set, generating a signal representing the prerequisite risk set, and storing the signal representing the prerequisite risk set in the memory. 45. The system of claim 24 further includes a classification module having processor-executable code and adapted to identify green categories and select entities for a set of entities included in the green categories, each having a green score, the selection being at least in part based on the respective entity's green score. 46. The system of claim 45, wherein the classification module includes processor-executable code to certify a company as green compliant, and wherein each of the selected entities is certified as green compliant.