HK1249598B - Method, apparatus and system for monitoring internet media events based on industry knowledge mapping database - Google Patents

Description

Method, device, and system for monitoring internet media events based on an industry knowledge graph database

Technical Field

The present invention relates to the field of Internet media monitoring, and in particular, to a technology for constructing an industry knowledge graph database and a technology for monitoring Internet media events based on the constructed industry knowledge graph database.

Background Art

The rapid development of computer, communications, and network technologies has continuously improved the performance of terminal devices, including PCs, tablets, smartphones, and Internet-connected televisions. Accordingly, internet media, particularly social media, has gradually become one of the primary channels for the public to obtain news and information, thanks to its diversity, speed, interactivity, ease of reproduction, and multimedia capabilities.

However, while internet media information offers advantages such as high timeliness and flexible and convenient access, the open nature of its information sources and dissemination methods also leads to the following problems: Without authorization or verification, sensitive information (e.g., trade secrets) or even false information can be rapidly disseminated by a large number of users on internet media platforms, potentially leading to media events that negatively impact individuals, businesses/institutions, industries, and even society. Therefore, it is necessary to monitor media events on the internet and, when certain conditions are met, take appropriate measures to reduce or eliminate their potential impact.

Existing Internet media monitoring technology has the following defects: 1) It uses interest matching to provide users with Internet media monitoring. Users need to customize the content topics and related entities they are interested in. Therefore, in the monitoring, only events directly related to the entities defined by the user can be identified, and events that are not defined by the user but are indirectly related to the entities of interest to the user cannot be identified; 2) The attributes of the monitored objects are single, and monitoring can only be provided for a single media category and data source (for example, specific social media, news media, forums, blogs, etc.), a single data type (generally text), and a single language.

Summary of the Invention

One purpose of the present invention is to provide a technology for constructing an industry knowledge graph database, extracting relevant data for a specific industry or field and storing it in the knowledge graph database. The constructed industry knowledge graph database can be applied to Internet media monitoring to achieve automated and in-depth monitoring of relevant Internet media events.

Another object of the present invention is to provide a technology for monitoring Internet media events based on the constructed industry knowledge graph database, which can identify non-directly related entities corresponding to specific media events during monitoring and can monitor various types of Internet media data.

In order to achieve the above-mentioned purpose of the invention, the specific technical solutions provided by the present invention are as follows.

The present invention provides a method for constructing an industry knowledge graph database, comprising the following steps: obtaining industry data from a data source; performing data processing on the industry data to extract entities related to the industry and corresponding entity attributes and/or entity relationships; and constructing the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the step of acquiring industry data is implemented by: acquiring structured industry data from a third-party industry database, wherein the structured industry data includes multiple fields; the step of processing the industry data is implemented by: performing data cleaning and extraction-transformation-loading (ETL) processing on the structured industry data; the step of constructing an industry knowledge graph database is implemented by: generating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the step of obtaining industry data is achieved by: using web crawler technology to obtain industry-related data from Internet data sources, and the Internet data sources include unstructured or semi-structured data sources; the step of processing industry data is achieved by: using information extraction technology in natural language processing to perform entity recognition and relationship extraction on the industry-related data to extract the entities, entity attributes and/or entity relationships; the step of constructing an industry knowledge graph database is achieved by: supplementing or updating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. Further preferably, the above steps are performed regularly at a predetermined period.

Preferably, the step of obtaining industry data is implemented by: obtaining industry-related data from an Internet data source in a query manner using an application programming interface (API), wherein the Internet data source includes an open data source; the step of processing the industry data is implemented by: before extracting entities related to the industry and corresponding entity attributes and/or entity relationships, performing data cleaning and extraction-transformation-loading (ETL) processing on the industry-related data; the step of constructing an industry knowledge graph database is implemented by: supplementing or updating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. Further preferably, the above steps are performed regularly at a predetermined period.

Preferably, the step of acquiring industry data is implemented by: utilizing an application programming interface (API) or web crawler technology to acquire industry-related internet media data from an internet data source; the step of processing the industry data is implemented by: performing event detection, event evaluation, and screening on the internet media data to extract specific media events related to the industry and identifying corresponding directly related entities from the internet media data; and the step of constructing an industry knowledge graph database is implemented by: supplementing the industry knowledge graph database based on the specific media events and the corresponding directly related entities, wherein the specific media events are supplemented into the industry knowledge graph database as abstract entities. Further preferably, in the step of processing the industry data, directly related entities corresponding to the specific media events are identified by at least one of the following methods: identifying entities from text data based on entity recognition in natural language processing; identifying entities from image or video data based on image or video recognition processing; or identifying entities from audio or video data based on speech recognition processing. Further preferably, the specific media events include negative events, emergencies, crisis events, mass incidents, public opinion events, or other events of industry significance. Further preferably, the above steps are performed in real time and continuously.

Preferably, the step of constructing the industry knowledge graph database includes: performing semantic disambiguation and entity linking on the extracted entities. Further preferably, the step of performing semantic disambiguation and entity linking on the extracted entities is further implemented by at least one of the following methods: performing semantic disambiguation and entity linking on each extracted entity reference independently based on entity knowledge; performing semantic disambiguation and entity linking on the extracted entity references consistently based on the subject consistency assumption and utilizing the associations of candidate entities in the knowledge base.

The present invention also provides a method for monitoring specific media events related to an industry based on the industry knowledge graph database constructed in the present invention, comprising the following steps: obtaining Internet media data; performing event detection, event evaluation and screening based on the obtained Internet media data to obtain the specific media events related to the industry; identifying directly related entities corresponding to the specific media events; based on the directly related entities, accessing the industry knowledge graph database to determine non-directly related entities corresponding to the specific media events; and sending early warning messages to the directly related entities and/or the non-directly related entities.

Preferably, the event detection in the event detection, event evaluation, and event screening steps includes the following steps: performing topic classification on the content in the acquired Internet media data to obtain content targeting a specific topic; identifying entities involved in the acquired content; performing sentiment analysis on the acquired content and the identified entities, and filtering the acquired content based on the results of the sentiment analysis; and performing event discovery based on the filtered content to cluster media events and discover new media events. Further preferably, the event detection also includes the following steps: analyzing the authenticity of the events based on the attributes of the media events, and sorting and/or filtering the media events based on the analysis results.

Preferably, in the step of identifying directly related entities corresponding to the specific media event, the directly related entities corresponding to the specific media event are identified by at least one of the following methods: identifying entities from text data based on entity recognition in natural language processing; identifying entities from image or video data based on image or video recognition processing; or, identifying entities from audio or video data based on speech recognition processing.

Preferably, the step of accessing the industry knowledge graph database is implemented in the following manner: based on the directly related entities, querying the industry knowledge graph database to determine the indirect related entities.

Preferably, the step of accessing the industry knowledge graph database is achieved by using data mining technology in the industry knowledge graph database based on the directly related entities to determine the non-directly related entities.

The present invention also provides a device for constructing an industry knowledge graph database, including: a data acquisition module for acquiring industry data from a data source; a data processing module for processing the industry data to extract entities related to the industry and corresponding entity attributes and/or entity relationships; a database construction module for constructing the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the data acquisition module acquires industry data in the following manner: obtaining structured industry data from a third-party industry database, the structured industry data including multiple fields; the data processing module performs data processing in the following manner: before extracting entities related to the industry and the corresponding entity attributes and/or entity relationships, the structured industry data is cleansed and subjected to extraction-transformation-loading (ETL) processing; the database construction module constructs an industry knowledge graph database in the following manner: generating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the data acquisition module acquires industry data in the following manner: using web crawler technology to obtain industry-related data from Internet data sources, and the Internet data sources include unstructured or semi-structured data sources; the data processing module performs data processing in the following manner: using information extraction technology in natural language processing to perform entity recognition and relationship extraction on the industry-related data to extract the entities, entity attributes and/or entity relationships; the database construction module constructs an industry knowledge graph database in the following manner: supplementing or updating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the data acquisition module acquires industry data in the following manner: using an application programming interface (API) to obtain industry-related data from an Internet data source in a query manner, and the Internet data source includes an open data source; the data processing module performs data processing in the following manner: before extracting entities related to the industry and corresponding entity attributes and/or entity relationships, the industry-related data is cleansed and subjected to extraction-transformation-loading (ETL) processing; the database construction module constructs an industry knowledge graph database in the following manner: supplementing or updating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the data acquisition module acquires industry data in the following manner: it is used to acquire industry-related Internet media data from Internet data sources using an application programming interface (API) or web crawler technology; the data processing module processes data in the following manner: it performs event detection, event evaluation and screening on the Internet media data to extract specific media events related to the industry, and identifies corresponding directly related entities from the Internet media data; the database construction module constructs an industry knowledge graph database in the following manner: based on the specific media events and the corresponding directly related entities, the industry knowledge graph database is supplemented, wherein the specific media events are supplemented into the industry knowledge graph database as abstract entities.

Preferably, the database construction module further identifies directly related entities corresponding to the specific media event in at least one of the following ways: identifying entities from text data based on entity recognition in natural language processing; identifying entities from image or video data based on image or video recognition processing; or identifying entities from audio or video data based on speech recognition processing.

Preferably, the database construction module includes: a module for performing semantic disambiguation and entity linking on the extracted entities. Further preferably, the module for performing semantic disambiguation and entity linking on the extracted entities further performs semantic disambiguation and entity linking in at least one of the following ways: performing semantic disambiguation and entity linking on each extracted entity reference independently based on entity knowledge; performing semantic disambiguation and entity linking on the extracted entity references consistently based on the subject consistency assumption and utilizing the associations of candidate entities in the knowledge base.

Preferably, the specific media events include negative events, emergencies, crisis events, mass events, public opinion events or other events with industry significance.

The present invention also provides a system for monitoring specific media events related to an industry, comprising: a data acquisition unit for obtaining industry data from a data source; a data processing unit for processing the industry data to extract entities related to the industry and corresponding entity attributes and/or entity relationships; a database construction unit for constructing the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships; a database storage unit for storing the constructed industry knowledge graph database; a media event monitoring unit for acquiring Internet media data, performing event detection, event evaluation and screening based on the acquired Internet media data to obtain the specific media events related to the industry, and identifying directly related entities corresponding to the specific media events; a database access unit for accessing the industry knowledge graph database based on the directly related entities to determine non-directly related entities corresponding to the specific media events; and a message sending unit for sending early warning messages to the directly related entities and/or the non-directly related entities.

Preferably, the data acquisition unit includes: a structured data acquisition unit for obtaining structured industry data from a third-party industry database, wherein the structured industry data includes multiple fields; the data processing unit includes: a structured data processing unit for performing data cleaning and extraction-transformation-loading (ETL) processing on the structured industry data before extracting entities related to the industry and corresponding entity attributes and/or entity relationships; the database construction unit includes: a database generation unit for generating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the data acquisition unit includes: an industry-related data acquisition unit, which is used to obtain industry-related data from Internet data sources using web crawler technology, and the Internet data sources include unstructured or semi-structured data sources; the data processing unit includes: an industry-related data processing unit, which is used to use information extraction technology in natural language processing to perform entity recognition and relationship extraction on the industry-related data to extract the entities, entity attributes and/or entity relationships; the database construction unit includes: a database supplement/update unit, which is used to supplement or update the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the data acquisition unit includes: an industry-related data acquisition unit, which is used to obtain industry-related data from an Internet data source in a query manner using an application programming interface (API), and the Internet data source includes an open data source; the data processing unit includes: an industry-related data processing unit, which is used to perform data cleaning and extraction-transformation-loading (ETL) processing on the industry-related data before extracting entities related to the industry and corresponding entity attributes and/or entity relationships; the database construction unit includes: a database supplement/update unit, which is used to supplement or update the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

Preferably, the data acquisition unit includes: a media data acquisition unit, which is used to obtain industry-related Internet media data from Internet data sources using an application programming interface (API) or web crawler technology; the data processing unit includes: a media data processing unit, which is used to perform event detection, event evaluation and screening on the Internet media data to extract specific media events related to the industry and identify corresponding directly related entities from the Internet media data; the database construction unit includes: a database supplement/update unit, which is used to supplement the industry knowledge graph database based on the specific media events and the corresponding directly related entities, wherein the specific media events are supplemented into the industry knowledge graph database as abstract entities.

Preferably, the database supplement/update unit is further used for: performing semantic disambiguation and entity linking on the extracted entities.

Preferably, the media event monitoring unit is further configured to: perform topic classification on the content of the acquired Internet media data to obtain content targeting specific topics; identify entities involved in the acquired content; perform sentiment analysis on the acquired content and the identified entities, and filter the acquired content based on the results of the sentiment analysis; and perform event discovery based on the filtered content to cluster media events and discover new media events. Further preferably, the media event monitoring unit is further configured to: analyze the authenticity of the events based on their attributes, and sort and/or filter the media events based on the analysis results.

Preferably, the database access unit is further used to: query the industry knowledge graph database based on the directly related entities to determine the indirect related entities.

Preferably, the database access unit is further used to: use data mining technology in the industry knowledge graph database based on the directly related entities to determine the indirect related entities.

Preferably, the specific media events include negative events, emergencies, crisis events, mass events, public opinion events or other events with industry significance.

By implementing the technical solution provided by the present invention, the following technical effects can be achieved: 1) Automated and in-depth monitoring of relevant Internet media events is achieved for one or more target fields or industries, and non-directly related entities corresponding to specific media events can be identified; 2) Automated processing of Internet media data from multiple data sources, multiple data types, and multiple languages is achieved during monitoring.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is an exemplary flow chart of a method for constructing an industry knowledge graph database provided by the present invention;

FIG2 is an exemplary structured industry data provided by the present invention;

FIG3 is an exemplary flow chart of a method for monitoring media events provided by the present invention;

FIG4 is an exemplary flow chart of another method for constructing an industry knowledge graph database provided by the present invention;

FIG5 is an exemplary flow chart of another method for constructing an industry knowledge graph database provided by the present invention;

FIG6 is an exemplary block diagram of a system for monitoring media events provided by the present invention.

DETAILED DESCRIPTION

The following describes the specific embodiments of the present invention in the form of embodiments in conjunction with the accompanying drawings to facilitate those skilled in the art to understand the objectives, technical solutions and advantages of the present invention. Those skilled in the art will understand that the specific embodiments described in the form of embodiments are merely exemplary and that the concept of the present invention can be implemented without these specific contents.

The present invention provides a technology for constructing an industry knowledge graph database and a technology for monitoring Internet media events based on the constructed industry knowledge graph database, so as to achieve the purpose of the present invention.

The present invention relates to the application of knowledge graph database technology. A knowledge graph database is a special database used for knowledge management, which facilitates the collection, organization and extraction of knowledge in related fields. Entities, entity attributes and entity relationships are defined in the knowledge graph database. Among them, entities correspond to things in the real world (for example, a company A, a person X), and each entity can be identified by a globally unique ID. Entity attributes are used to describe the intrinsic characteristics of an entity (for example, the Chinese and English names of company A and person X). Entity relationships are used to connect entities to describe the connection between entities (for example, the position relationship between person X and company A). By constructing a knowledge graph database, knowledge composed of entities, entity attributes and entity relationships can be more efficiently and deeply utilized to discover complex connections between things.

As a database, a knowledge graph database can be stored in a variety of forms. For example, a knowledge graph database can be stored in a traditional relational database using semantic network RDF (Resource Description Framework) triples, or it can be stored in a new non-relational database. Preferably, the knowledge graph database can be stored in a graph database, such as Neo4j, OrientDB, Titan-BerkeleyDB, HyperGraphDB, etc.

Depending on the scale and purpose of the knowledge graph database, the data sources used to build the knowledge graph database can be diverse. For example, the data source can be an open encyclopedia data source (for example, Baidu Encyclopedia, Wikipedia, etc.), a structured database (for example, Wikidata, DBpedia, vertical websites or professional databases for specific industries, etc.), or any relevant third-party semi-structured or unstructured data source (for example, professional websites, content published in online media, including news, company annual reports, corporate announcements, etc.).

Those skilled in the art will appreciate that the knowledge graph database constructed in this invention is oriented toward specific fields or industries, but is not limited to a single industry. The constructed knowledge graph database integrates and connects entities and events related to one or more industries, their attributes, and the relationships between entities, entities and events, and events and events into a knowledge graph.

FIG1 is an exemplary flowchart of a method for constructing an industry knowledge graph database provided by the present invention, which may include steps S11-S15.

In step S11, industry data is obtained from an industry data source, and entities and corresponding entity attributes and entity relationships are extracted from the industry data to generate the industry knowledge graph database.

An industry data source is a source of basic data for one or more specific fields or industries targeted for monitoring. In one embodiment, the industry data source can be a structured industry database to obtain high-quality basic industry data. The structured database can be accessed through an application programming interface (API) to obtain data through queries (e.g., through query commands).

Through the "Extraction-Transform-Load (ETL)" process, the obtained industry data can be transformed, and then entities, entity attributes and entity relationships can be extracted from the transformed data and loaded into the industry knowledge graph database proposed by the present invention. The specific execution steps of the ETL operation can be achieved through existing data integration means. For example, in the ontology-based data integration method, the mapping relationship between each field in different databases and various entity information is defined in a predetermined manner, so that entities, entity attributes and entity relationships are extracted according to the fields and their contents, and the basic industry knowledge graph database is constructed. In addition, since industry databases are different in structure and may have problems such as data noise, data missing or data errors, it may be necessary to perform data cleaning operations on the industry data during the process of data processing. Technical means known in the art can be used in combination with ETL processing to achieve data cleaning operations.

As an example, FIG2 shows exemplary structured industry data, which, as described above, can be obtained from a structured industry database. In FIG2 , Table 1 is an example of structured data of a listed company, which includes two data entries, Company A and Company B, and each data entry includes multiple fields such as the company's Chinese and English names, registered address, stock code, and chairman of the board of directors. By performing ETL operations on the structured data, the entities (i.e., Company A, Company B, Person X, Person Y), entity attributes (i.e., specific information about Company A and Company B), and entity relationships (i.e., the relationship between Company A and Person X, and between Company B and Person Y) can be extracted, thereby generating a knowledge graph database for the industry to which it belongs.

In another embodiment, the industry data source may also be a semi-structured or unstructured data source from the Internet, and industry data may be captured from the data source through web crawler technology, and information extraction operations based on natural language processing technology may be used to extract entities, entity attributes, and entity relationships.

In step S12, data related to the industry is obtained from an Internet data source, and entities related to the industry and corresponding entity attributes and entity relationships are extracted from the data.

In this step, data related to the aforementioned specific domain or industry is first obtained from internet data sources. Internet data sources can be structured, semi-structured, or unstructured. Therefore, different approaches can be used to obtain industry-related data based on the different structural characteristics of internet data sources. Entities, their corresponding entity attributes, and entity relationships are then extracted from this industry-related data.

For structured Internet data sources, you can query the corresponding data content through the API and obtain entities, entity attributes and entity relationships. For semi-structured data sources, after crawling the data content, you can analyze the content through information extraction operations in natural language processing technology to extract industry-related entities, entity attributes and entity relationships. A semi-structured data source is a data source that contains partially structured and partially unstructured data. Therefore, the corresponding parts of the semi-structured data can be processed in the same way as structured and unstructured data. For example, HTML and XML files are the most common semi-structured data. In the process of processing HTML and XML files, on the one hand, you can use the tag-based structured information therein, and on the other hand, you can combine information extraction technology with machine learning technology to extract the required information.

In one embodiment, the information extraction operation includes an entity recognition operation and a relationship extraction operation.

Entity recognition operations can use existing natural language processing tools (for example, part-of-speech tagging or named entity recognition tools), or use machine learning methods to train entity recognition models on specific annotated data. It should be noted that some natural language processing tasks and processing tools are language-dependent (for example, Chinese data requires word segmentation, while English data does not). Machine learning methods digitally represent data in different languages and formats, and then use general, language-independent algorithms (for example, conditional random field algorithms and hidden Markov models) for model training.

Relationship extraction can be achieved through a variety of existing statistical learning or machine learning methods. For example, a template learning approach can be used, using entities in a knowledge graph database that match a certain relationship as instances. The existing instances are extracted from a large amount of text and statistically analyzed for sentence patterns and contexts, forming a relationship extraction template. This template can then be applied to the text data to extract new instances. If an instance is extracted that does not yet exist in the knowledge graph database, it can be added to the knowledge graph database.

In step S13, the industry knowledge graph database is supplemented or updated based on the industry-related entities and the corresponding entity attributes and entity relationships.

After extracting industry-related entities and corresponding entity attributes and entity relationships, they can be associated and compared with the corresponding information in the knowledge graph database, and new entities, entity attributes and entity relationships can be added to the knowledge graph database as needed, and existing entity attributes and entity relationships can be updated.

As described above, the industry knowledge graph database proposed in the present invention can adopt a traditional relational database, an RDF triple database, or a new non-relational database (e.g., a graph database). Correspondingly, the specific operation of supplementing or updating the knowledge graph database can be implemented in a customized manner using a database query language, such as SQL for relational databases, SPARQL for RDF triples, and Cypher for Neo4j graph databases.

Continuing with the example in Figure 2, let's assume that the structured data on listed company executives in Table 2 is obtained from a structured internet data source via an API query. The following additions and updates can be made to the industry knowledge graph database: 1) Person Z, Person Z's entity attributes, and the employment relationship between Person Z and Company B are added to the knowledge graph database; 2) Person X and Person Y's entity attributes are added; and 3) the employment relationship between Person Y and Company B is updated (i.e., from "currently employed" to "previously employed").

In one embodiment, entity linking operations and semantic disambiguation operations are required in the process of supplementing or updating the industry knowledge graph database.

Entity linking aims to map an entity reference (or entity mention) in the data content to a related entity concept in the knowledge graph database. For example, in the sentences "Jobs is one of the founders of Apple" and "Steve Jobs founded NeXT in the United States in 1985," both entity references "Jobs" and "Steve Jobs" should map to the same person entity concept in the knowledge graph database, "Steve Jobs (ex-CEO of Apple)." Therefore, entity linking is needed to link these two entity references to the same entity. Semantic disambiguation aims to resolve ambiguous entity references. For example, the entity reference "apple" can map to multiple ambiguous entities, such as "apple (fruit)", "Apple Inc.", and "apple (movie)." However, in the first sentence above, "apple" should map to the company entity concept "Apple Inc." in the knowledge graph database, not "apple (fruit)" or "apple (movie)." Entity linking and semantic disambiguation are typically performed together. Because semantic disambiguation is a means of entity linking, and entity linking is the goal of semantic disambiguation; the two are often used interchangeably or represent each other in different situations.

Any existing entity linking and semantic disambiguation technology can be used in the present invention. For example, one type of method is based on entity knowledge to independently disambiguate and link entity references one by one. Entity knowledge includes, but is not limited to, the probability of occurrence of the entity, the name distribution of the entity (full name, alias, abbreviation, etc.), the context of the entity (such as word co-occurrence information, word distribution, etc.), and the category information of the entity in the knowledge base (such as company entity, personal entity, location entity, etc.). Probabilistic methods (such as linear regression or logistic regression) or machine learning methods (such as support vector machines (SVMs) and random forests (Random Forests)) can be used to learn and train semantic disambiguation and entity linking models based on entity knowledge. Another type of method is based on the assumption of topic consistency (i.e., entities in an article are usually related to the text topic, so these entities also have semantic relevance). It uses the association of candidate entities of all entity references in the text content in the knowledge base (such as Wikipedia or the knowledge graph constructed by the present invention) to consistently disambiguate and link all entity references in an article. This type of method typically uses collaborative reasoning based on graph data structures during computation. Specifically, they construct a candidate entity graph based on the relationships between all entity references in the article's content. The dense distribution of the graph reflects the degree of semantic association between different candidate entity nodes. The entity linking process involves iteratively propagating evidence (the potential associations between different entities) along the dependency structure of the candidate entity graph to collaboratively enhance the evidence until convergence. The two aforementioned methods can also be flexibly and organically combined to improve disambiguation and linking performance.

In step S14, Internet media data related to the industry is obtained from an Internet data source, and specific media events related to the industry and corresponding directly related entities are extracted from the Internet media data.

There are many ways to obtain internet media data from internet data sources. For example, some social media sites (such as Sina Weibo, Facebook, and Twitter) have open APIs for accessing their data. Web crawling and content extraction techniques can also be used to scrape data from news websites or industry media websites.

Various technical implementations exist in the field for monitoring internet media for specific media events. For example, in one implementation, internet media data is first monitored to identify the content of media events within a specific field or industry of interest, as well as the entities involved in the events. These newly discovered media events are then evaluated based on various metrics (e.g., negativity, significance, suddenness, speed and scope of dissemination, credibility, etc.) to select eligible media events.

Different processing techniques can be used to identify directly related entities corresponding to media events for different types of Internet media data. For example, entity recognition techniques based on natural language processing can be used to identify entities from text data, image or video recognition processing techniques can be used to identify entities from image or video data, and speech recognition processing techniques can be used to identify entities from audio or video data. Those skilled in the art will appreciate that the present invention is not limited to the media type or language of Internet media data.

In step S15, the industry knowledge graph database is supplemented based on the specific media event and the corresponding directly related entities, wherein the specific media event is supplemented into the industry knowledge graph database as an abstract entity.

After obtaining a specific media event related to the industry and the corresponding directly related entities (for example, a corruption scandal of the chairman of a listed company and the companies, people, and places involved in the incident), the event is added to the industry knowledge graph database as an abstract entity, and at the same time, entity linking and semantic disambiguation are performed on the directly related entities involved in the event, that is, the entity corresponding to the entity in the industry knowledge graph database is found, and it is associated with the abstract entity representing the event. If it is found that the entity involved in the event does not exist in the industry knowledge graph database, it can be supplemented in the manner described in step S13 above. After completing the supplement to the industry knowledge graph database, other non-directly related entities of the abstract entity representing the media event in the industry knowledge graph database can be found based on the relationship between the directly related entities of the event and other entities in the knowledge graph database.

After constructing the industry knowledge graph database in the above manner, it is possible to conduct automated, in-depth monitoring of Internet media events based on the constructed information. Preferably, after the initial construction of the industry knowledge graph database is completed, in order to maintain the integrity and validity of the information, the industry knowledge graph database can also be updated. For example, steps S12 and S13 can be performed regularly at a predetermined period, or steps S14 and S15 can be performed in real time and continuously.

In addition, those skilled in the art will understand that the content of various data such as industry data, industry-related data, and Internet media data involved in the present invention can be in multiple languages and can also be of multiple types (for example, text, images, videos, voice, etc.), and the present invention does not impose any restrictions on this.

Figure 3 is an exemplary flow chart of a method for monitoring media events provided by the present invention, which can monitor specific media events related to an industry based on the industry knowledge graph database constructed in the present invention. The method can include steps S31-S35.

In step S31 , Internet media data is acquired.

As mentioned above, there are many ways to obtain internet media data from internet data sources. For example, some social media sites (such as Sina Weibo, Facebook, and Twitter) have open APIs for accessing their data. Web crawling and content extraction techniques can also be used to scrape data from news websites or industry media websites.

In step S32, event detection, event evaluation and screening are performed based on the acquired Internet media data to obtain the specific media events related to the industry.

As described above, various technical implementations exist in the art for monitoring internet media to identify specific media events. For example, in one implementation, internet media data is first monitored to identify the content of media events within a specific field or industry of interest, as well as the entities involved in the events. These newly discovered media events are then evaluated based on various metrics (e.g., negativity, significance, suddenness, speed and scope of dissemination, credibility, etc.) to select eligible media events.

Specifically, in one embodiment, the technical implementation steps involved in event detection may include: topic classification, entity recognition, sentiment analysis, and event discovery.

In the topic classification step, the content in the acquired Internet media data is subject to topic classification to obtain content for a specific topic. The purpose of topic classification is to filter out text belonging to a certain topic of interest or a category related to customer needs from the acquired content. Topic classification is a text mining technology that generally uses machine learning or deep learning methods to train a classification model on labeled data, and then applies it to the text to determine its topic category. Any existing classification model (for example, a naive Bayes model, a decision tree, a support vector machine, an artificial neural network, etc.) can be used in the present invention.

During the entity recognition step, entities involved are identified from the acquired content. The purpose of entity extraction is to identify entities involved in the article for further analysis. For example, entity recognition can include extracting entities from text using information extraction techniques from natural language processing, identifying entities from images (including videos) using image recognition techniques, and identifying entities from speech using speech recognition techniques. Entities identified from text, images, and speech can also be combined.

In the sentiment analysis step, sentiment analysis is performed on the obtained content and the identified entities, and the obtained content is filtered based on the results of the sentiment analysis. Sentiment analysis is used to determine the full text of the content and the polarity of the sentiment expressed for different entities in order to find content that meets the monitoring conditions. Existing technologies generally implement sentiment analysis using text classification methods (for example, classifying sentiment as positive, neutral, or negative) or regression analysis methods (for example, expressing sentiment as a score between -5 and +5). To determine the sentiment towards a certain entity in the content, the contextual information of the entity in the text can be used, or dependency syntactic analysis tools can be used to find the text portion in the text that is related to the entity to perform sentiment analysis on the entity.

In the event discovery step, event discovery is performed based on the filtered content to cluster media events and discover new media events. The goal of event discovery is to extract event information (e.g., the time and location of the event) from different texts. This information is then clustered and merged into abstract "events." New events are identified by comparing them with existing events, and events are clustered based on content similarity or relevance.

In one embodiment, optionally, during the event detection process, the authenticity of the event can also be analyzed based on the attributes of the media event (for example, the time and place of the event, the publisher of the media event and its related attributes, etc.), and the media events can be sorted and/or filtered according to the analysis results.

Those skilled in the art will understand that the implementation methods listed for each operation in the above steps are merely exemplary, and some other existing methods in the art can also implement these operations. The present invention does not impose any limitation on the specific methods for implementing the above operations.

In step S33 , directly related entities corresponding to the specific media event are identified.

In one embodiment, through entity recognition and event discovery operations in event monitoring, directly related entities in each media event can be obtained. At the same time, as described above, each directly related entity can be associated with a corresponding entity concept in the industry knowledge graph database or added to the industry knowledge graph database through entity linking and semantic disambiguation.

In step S34, based on the directly related entities, the industry knowledge graph database is accessed to determine indirect related entities corresponding to the specific media event.

In one embodiment, other non-directly related entities that have an association relationship with the entity directly related to the event can be directly queried on the industry knowledge graph database through various preset conditions. For example, the preset conditions can be: 1) entities that have an association relationship with the entity directly related to the event within N layers (N can be 1, 2, 3...); 2) other entities whose association degree with the entity directly related to the event meets certain conditions (such as being greater than a specified threshold); 3) entities that have a certain specific relationship with the entity directly related to the event (for example, a supply relationship, an investment relationship, etc.); 4) entities with certain specific attributes (for example, belonging to a specified industry, being located in a certain location, holding a certain position, etc.). These preset conditions can be used individually or in any combination.

In another embodiment, a data mining method can be used to mine non-directly related entities of an event using multiple conditions based on an industry knowledge graph database. For example, a specific implementation method can use link prediction technology for graph data, that is, the problem of detecting non-directly related entities of an event can be expressed as the technical problem of "predicting whether there is an edge between the node representing the event and other entity nodes other than the directly related entity nodes in the industry knowledge graph database." The conditions that can be used for link prediction include but are not limited to the characteristics of the event itself (for example, the type of event, time and place attributes, negativity, etc.), the relationship between the event and historical events (including the type of relationship and the strength of the relationship), the relationship between the directly related entity of the event and other entities (including the type of relationship and the strength of the relationship), and all knowledge that can be mined in the knowledge graph database, such as entity types and attributes, thereby achieving a comprehensive judgment of the non-directly related entities of a specific media event.

In step S35, an early warning message is sent to the directly related entity and/or the indirect related entity.

After identifying directly and indirectly related entities corresponding to a specific media event, warning messages can be sent to the corresponding entity users through various channels (e.g., email, mobile text messages, real-time chat tools, social networking platforms, etc.). Warning messages can include a text description of the event itself, images, relevant dissemination statistics, event evaluation indicators, and how related entities may be affected by the event.

Those skilled in the art will appreciate that the specific media events described in the present invention may be various types of events that meet user-defined conditions and can be obtained from Internet media, such as negative events, emergencies, crisis events, mass incidents, or public opinion events. The present invention does not impose any limitation thereto.

As a preferred embodiment, Figure 4 shows an exemplary flow chart of another method for constructing an industry knowledge graph database provided by the present invention. The method may include steps S41, S421/S422, and S43-S45.

In step S41, industry data is obtained from an industry data source, and entities and corresponding entity attributes and entity relationships are extracted from the industry data to generate an industry knowledge graph database.

In step S421, entities, entity attributes, and entity relationships related to the industry are obtained by querying the structured data source using an application programming interface. In one embodiment, the structured data source can be a structured open data platform such as Wikidata or DBPedia, and industry-related data can be obtained from it through an API.

In step S422, entity recognition and relationship extraction are performed on the data using natural language processing techniques based on the semi-structured or unstructured data source to extract entities, entity attributes, and entity relationships related to the industry. In one embodiment, the semi-structured or unstructured data source can be an open data platform such as Wikipedia or Baidu Encyclopedia, or any relevant third-party data source (e.g., professional websites, content published in internet media, etc.). Industry-related data can be obtained through web crawling or content extraction techniques.

Preferably, steps S421 and/or S422, S43 may be performed regularly at a predetermined cycle.

In step S43, the industry knowledge graph database is supplemented or updated based on the industry-related entities and the corresponding entity attributes and entity relationships.

In step S44, Internet media data is obtained from an Internet data source, and specific media events related to the industry and corresponding directly related entities are extracted from the Internet media data.

In step S45, the industry knowledge graph database is supplemented based on the specific media event and the corresponding directly related entities, wherein the specific media event is supplemented into the industry knowledge graph database as an abstract entity.

Preferably, steps S44 and S45 can be performed in real time and continuously.

FIG5 is an exemplary flow chart of another method for constructing an industry knowledge graph database provided by the present invention. The method may include steps S51-S53:

In step S51, industry data is obtained from a data source;

In step S52, the industry data is processed to extract entities related to the industry and corresponding entity attributes and/or entity relationships;

In step S53, the industry knowledge graph database is constructed based on the extracted entities, entity attributes and/or entity relationships.

As mentioned above, the data sources for industry knowledge graph databases can be diverse, including but not limited to open encyclopedia data sources, structured databases, and any relevant third-party semi-structured or unstructured internet data sources. Furthermore, as mentioned above, the data sources for industry knowledge graph databases can also be internet media data sources.

In one embodiment, the data source may be a structured industry database, and the method may be implemented in the following specific manner: in step S51(1), structured industry data including multiple fields is obtained from a third-party industry database; in step S52(1), before extracting entities related to the industry and corresponding entity attributes and/or entity relationships, the structured industry data is cleansed and subjected to extraction-transformation-loading (ETL) processing; in step S53(1), the industry knowledge graph database is generated based on the extracted entities, entity attributes and/or entity relationships.

In another embodiment, the data source may be an unstructured or semi-structured Internet data source, and the method may be implemented in the following specific manner: in step S51(2), using web crawler technology, industry-related data is obtained from an Internet data source, and the Internet data source includes an unstructured or semi-structured data source; in step S52(2), using information extraction technology in natural language processing, entity recognition and relationship extraction are performed on the industry-related data to extract the entities, entity attributes and/or entity relationships; in step S53(2), the industry knowledge graph database is supplemented or updated based on the extracted entities, entity attributes and/or entity relationships.

In addition, the steps S51(2)-S53(2) may be performed regularly at a predetermined cycle.

In another embodiment, the data source may be an open Internet data source, and the method may be implemented in the following specific manner: in step S51(3), industry-related data is obtained from the Internet data source in a query manner using an application programming interface (API); in step S52(3), before extracting entities related to the industry and corresponding entity attributes and/or entity relationships, the industry-related data is cleansed and subjected to extraction-transformation-loading (ETL) processing; in step S53(3), the industry knowledge graph database is supplemented or updated based on the extracted entities, entity attributes and/or entity relationships.

In addition, the steps S51(3)-S53(3) may be performed regularly at a predetermined cycle.

In another embodiment, the data source may be an Internet media data source, and the method may be implemented in the following specific manner: in step S51(4), Internet media data is obtained from an Internet data source using an application programming interface (API) or web crawler technology; in step S52(4), event detection, event evaluation, and screening are performed on the Internet media data to extract specific media events related to the industry, and corresponding directly related entities are identified from the Internet media data; in step S53(4), the industry knowledge graph database is supplemented based on the specific media events and the corresponding directly related entities, wherein the specific media events are supplemented into the industry knowledge graph database as abstract entities.

For example, in step S52(4), directly related entities corresponding to the specific media event can be identified by at least one of the following methods: identifying entities from text data based on entity recognition in natural language processing; identifying entities from image or video data based on image or video recognition processing; or identifying entities from audio or video data based on speech recognition processing.

For example, the specific media events may include negative events, emergencies, crisis events, mass events, public opinion events, or other events with industry significance.

In addition, steps S51(4)-S53(4) can be performed in real time and continuously.

In another embodiment, the step of supplementing or updating the industry knowledge graph database in the above steps S53(2), S53(3), and S53(4) may include: performing semantic disambiguation and entity linking on the extracted entities. For example, the semantic disambiguation and entity linking may be performed in at least one of the following ways: based on entity knowledge, performing semantic disambiguation and entity linking on each extracted entity reference independently; based on the topic consistency assumption, utilizing the association of candidate entities in the knowledge base, performing semantic disambiguation and entity linking on the extracted entity references consistently.

The above describes a method for constructing an industry knowledge graph database provided by the present invention in the form of embodiments. Those skilled in the art will understand that various combinations of these embodiments are also included in the concept of this method for constructing an industry knowledge graph database.

Figure 6 is an exemplary block diagram of a system for monitoring media events provided by the present invention. The system includes a data acquisition unit, a data acquisition unit, a database construction unit, a database storage unit, a media event monitoring unit, a database access unit, and a message sending unit.

The data acquisition unit is used to obtain industry data from the data source.

a data processing unit, configured to process the industry data to extract entities related to the industry and corresponding entity attributes and/or entity relationships;

A database construction unit, configured to construct the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships;

Database storage unit: used to store the constructed industry knowledge graph database;

A media event monitoring unit is configured to acquire Internet media data, perform event detection, event evaluation, and screening based on the acquired Internet media data to obtain the specific media events related to the industry, and identify directly related entities corresponding to the specific media events;

A database access unit is configured to access the industry knowledge graph database based on the directly related entities to determine the indirect related entities corresponding to the specific media event;

A message sending unit is used to send a warning message to the directly related entity and/or the indirect related entity.

In one embodiment, the data acquisition unit includes: a structured data acquisition unit for obtaining structured data from a third-party industry database, wherein the structured data includes multiple fields; the data processing unit includes: a structured data processing unit for performing data cleaning and extract-transform-load (ETL) processing on the structured data; the database construction unit includes: a database generation unit for generating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

In another embodiment, the data acquisition unit includes: an industry-related data acquisition unit, which is used to obtain industry-related data from Internet data sources using web crawler technology, and the Internet data sources include unstructured or semi-structured data sources; the data processing unit includes: an industry-related data processing unit, which is used to use information extraction technology in natural language processing to perform entity recognition and relationship extraction on the industry-related data to extract the entities, entity attributes and/or entity relationships; the database construction unit includes: a database supplement/update unit, which is used to supplement or update the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

In another embodiment, the data acquisition unit includes: an industry-related data acquisition unit, which is used to obtain industry-related data from an Internet data source in a query manner using an application programming interface (API), and the Internet data source includes an open data source; the data processing unit includes: an industry-related data processing unit, which is used to perform data cleaning and extraction-transformation-loading (ETL) processing on the industry-related data before extracting entities related to the industry and corresponding entity attributes and/or entity relationships; the database construction unit includes: a database supplement/update unit, which is used to supplement or update the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships.

In another embodiment, the data acquisition unit includes: a media data acquisition unit, which is used to obtain industry-related Internet media data from Internet data sources using an application programming interface (API) or web crawler technology; the data processing unit includes: a media data processing unit, which is used to perform event detection, event evaluation and screening on the Internet media data to extract specific media events related to the industry and identify corresponding directly related entities from the Internet media data; the database construction unit includes: a database supplement/update unit, which is used to supplement the industry knowledge graph database based on the specific media events and the corresponding directly related entities, wherein the specific media events are supplemented into the industry knowledge graph database as abstract entities.

In one embodiment, the database supplement/update unit is further configured to perform semantic disambiguation and entity linking on the extracted entities.

In one embodiment, the media event monitoring unit is further configured to: perform topic classification on the content in the acquired Internet media data to obtain content targeting a specific topic; identify entities involved in the acquired content; perform sentiment analysis on the acquired content and the identified entities, and filter the acquired content based on the results of the sentiment analysis; and perform event discovery based on the filtered content to cluster media events and discover new media events. In another embodiment, the media event monitoring unit is further configured to: analyze the authenticity of the events based on their attributes, and sort and/or filter the media events based on the analysis results.

In one embodiment, the database access unit is further configured to: query the industry knowledge graph database based on the directly related entities to determine the indirect related entities. In another embodiment, the database access unit is further configured to: use data mining technology in the industry knowledge graph database based on the directly related entities to determine the indirect related entities.

The above describes a system for monitoring media events provided by the present invention by way of example. Those skilled in the art will appreciate that the operating steps in the various methods described above in conjunction with Figures 1 and 3-5 can be applied to the components of the system, and therefore will not be described in detail here.

Those skilled in the art will also appreciate that the various exemplary method steps and units described in conjunction with the various embodiments disclosed herein can all be implemented as electronic hardware, computer software, or a combination thereof. In order to clearly illustrate the interchangeability of hardware and software, the various exemplary steps and units hereinabove have been generally described around their functions. Whether such functions are implemented as hardware or software depends on the specific application and the design constraints imposed on the entire system. Those skilled in the art may implement the described functions in an adaptable manner for each specific application, but such implementation decisions should not be interpreted as causing a departure from the scope of this disclosure.

The term "exemplary" used in the present invention description means used as an example, illustration or description. Any technical solution described as "exemplary" in the description should not be interpreted as being more preferred or more advantageous than other technical solutions.

The present invention provides the above description of the disclosed technical content to enable those skilled in the art to implement or use the present invention. Many modifications and variations of these technical contents are obvious to those skilled in the art, and the overall principles defined in the present invention can also be applied to other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the specific embodiments shown above, but should be consistent with the widest range of inventive concepts disclosed in the present invention.

Claims (38)

1. A method for constructing an industry knowledge graph database, characterized by comprising the following steps: Step 101: Obtain industry data of one or more specific industries from a data source, wherein the industry data includes Internet media data; Step 102: Process the industry data to extract entities related to the industry and corresponding entity attributes and/or entity relationships, and perform event detection, event evaluation, and screening on the Internet media data to extract specific media events related to the industry, wherein event detection includes: performing topic classification on the Internet media data to obtain content for a specific topic; identifying entities involved in the obtained content; performing sentiment analysis on the obtained content and the identified entities, and filtering the obtained content based on the results of the sentiment analysis; and performing event discovery based on the filtered content to cluster media events and discover new media events. Step 103: perform semantic disambiguation and entity linking on the extracted entities, and construct the industry knowledge graph database based on the extracted specific media events and the entities, entity attributes and/or entity relationships. 2. The method according to claim 1, characterized in that The step 101 further includes: acquiring structured industry data from a third-party industry database, wherein the structured industry data includes a plurality of fields; The step 102 further includes: before extracting entities related to the industry and corresponding entity attributes and/or entity relationships, performing data cleaning and extraction-transformation-loading (ETL) processing on the structured industry data; The step 103 also includes: generating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 3. The method according to claim 1, characterized in that The step 101 further includes: using web crawler technology to obtain industry-related data from an Internet data source, wherein the Internet data source includes an unstructured or semi-structured data source; The step 102 further includes: using information extraction technology in natural language processing to perform entity recognition and relationship extraction on the industry-related data to extract the entities, entity attributes and/or entity relationships; The step 103 also includes: supplementing or updating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 4. The method according to claim 1, wherein The step 101 further includes: obtaining industry-related data from an Internet data source in a query manner using an application programming interface (API), wherein the Internet data source includes an open data source; The step 102 further includes: before extracting entities related to the industry and corresponding entity attributes and/or entity relationships, performing data cleaning and Extract-Transform-Load (ETL) processing on the industry-related data; The step 103 also includes: supplementing or updating the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 5. The method according to claim 1, wherein The step 101 further includes: using an application program interface (API) or web crawler technology to obtain industry-related Internet media data from an Internet data source; The step 102 further includes: identifying directly related entities corresponding to the specific media event from the Internet media data; The step 103 further includes: supplementing the industry knowledge graph database based on the specific media event and the corresponding directly related entity, wherein the specific media event is supplemented into the industry knowledge graph database as an abstract entity. 6. The method according to claim 5, wherein in step 102, directly related entities corresponding to the specific media event are further extracted by at least one of the following methods: Identify entities from text data based on entity recognition in natural language processing; identifying entities from image or video data based on image or video recognition processing; or Identify entities from audio or video data based on speech recognition processing. 7. The method according to any one of claims 3-5 is characterized in that step 103 includes: performing semantic disambiguation and entity linking on the extracted entities to identify corresponding entities of the entities in the industry knowledge graph database. 8. The method according to claim 7, wherein the step of performing semantic disambiguation and entity linking on the extracted entities is further implemented by at least one of the following methods: Based on entity knowledge, semantic disambiguation and entity linking are performed independently for each extracted entity reference. Based on the topic consistency assumption, the associations of candidate entities in the knowledge base are utilized to perform consistent semantic disambiguation and entity linking on the extracted entity references. 9. The method according to claim 1, wherein the specific media event includes a negative event, an emergency event, a crisis event, a mass event or a public opinion event. 10. The method according to claim 3 or 4, characterized in that steps 101-103 are performed regularly in a predetermined cycle. 11. The method according to claim 5, characterized in that steps 101-103 are performed in real time and continuously. 12. A method for monitoring specific media events related to an industry using an industry knowledge graph database constructed based on the method of any one of claims 1 to 11, characterized by comprising the following steps: Step 1201, obtaining Internet media data; Step 1202: performing event detection, event evaluation, and screening based on the acquired Internet media data to obtain the specific media events related to the industry; Step 1203: extracting directly related entities corresponding to the specific media event; Step 1204: determining an indirect related entity corresponding to the specific media event based on the association relationship of the directly related entity in the industry knowledge graph database; Step 1205: Send an early warning message to the directly related entity and/or the indirect related entity. 13. The method according to claim 12, wherein the event detection in step 1202 comprises the following steps: Topic classification of the content in the acquired Internet media data to obtain content targeting specific topics; identifying involved entities from the obtained content; Performing sentiment analysis on the obtained content and the identified entities, and filtering the obtained content based on the result of the sentiment analysis; Event discovery is performed based on the filtered content to cluster media events and discover new media events. 14. The method according to claim 13, wherein the event detection in step 1202 further comprises the following steps: The authenticity of the media events is analyzed based on their attributes, and the media events are sorted and/or filtered according to the analysis results. 15. The method according to claim 12, wherein in step 1203, directly related entities corresponding to the specific media event are extracted by at least one of the following methods: Identify entities from text data based on entity recognition in natural language processing; identifying entities from image or video data based on image or video recognition processing; or Identify entities from audio or video data based on speech recognition processing. 16. The method according to claim 12, wherein step 1204 is implemented by: Semantic disambiguation and entity linking are performed on the directly related entities to identify corresponding entities in the industry knowledge graph database, and based on the directly related entities, other entities and/or entity relationships in the industry knowledge graph database are queried under preset conditions to determine the non-directly related entities. 17. The method according to claim 12, wherein step 1204 is implemented by: Semantic disambiguation and entity linking are performed on the directly related entities to identify corresponding entities in the industry knowledge graph database, and based on the directly related entities, data mining technology is used on other entities and/or entity relationships in the industry knowledge graph database to determine the non-directly related entities. 18. The method according to any one of claims 16-17, wherein the step of performing semantic disambiguation and entity linking on directly related entities is further implemented by at least one of the following means: Based on entity knowledge, semantic disambiguation and entity linking are performed independently for each extracted entity reference. Based on the topic consistency assumption, the associations of candidate entities in the knowledge base are utilized to perform consistent semantic disambiguation and entity linking on the extracted entity references. 19. A device for constructing an industry knowledge graph database, comprising: A data acquisition module, configured to acquire industry data of one or more specific industries from a data source, wherein the industry data includes Internet media data; A data processing module is used to process the industry data to extract entities related to the industry and corresponding entity attributes and/or entity relationships, and to perform event detection, event evaluation and screening on the Internet media data to extract specific media events related to the industry, wherein event detection includes: performing topic classification on the Internet media data to obtain content for a specific topic; identifying entities involved from the obtained content; performing sentiment analysis on the obtained content and the identified entities, and filtering the obtained content based on the results of the sentiment analysis; and performing event discovery based on the filtered content to cluster media events and discover new media events. A database construction module is used to perform semantic disambiguation and entity linking on the extracted entities, and to construct the industry knowledge graph database based on the extracted specific media events and the entities, entity attributes and/or entity relationships. 20. The device according to claim 19, characterized in that The data acquisition module is further used to: obtain structured industry data from a third-party industry database, wherein the structured industry data includes multiple fields; The data processing module is further configured to: perform data cleaning and Extract-Transform-Load (ETL) processing on the structured industry data before extracting entities related to the industry and corresponding entity attributes and/or entity relationships; The database construction module is also used to generate the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 21. The device according to claim 19, characterized in that The data acquisition module is further configured to: utilize web crawler technology to obtain industry-related data from Internet data sources, wherein the Internet data sources include unstructured or semi-structured data sources; The data processing module is further configured to: utilize information extraction technology in natural language processing to perform entity recognition and relationship extraction on the industry-related data to extract the entities, entity attributes and/or entity relationships; The database construction module is also used to supplement or update the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 22. The device according to claim 19, characterized in that The data acquisition module is further configured to: acquire industry-related data from an Internet data source in a query manner using an application programming interface (API), wherein the Internet data source includes an open data source; The data processing module is further configured to: perform data cleaning and Extract-Transform-Load (ETL) processing on the industry-related data before extracting entities and corresponding entity attributes and/or entity relationships related to the industry; The database construction module is also used to supplement or update the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 23. The device according to claim 19, characterized in that The data acquisition module is further used to: acquire industry-related Internet media data from Internet data sources using an application programming interface (API) or web crawler technology; The data processing module is further configured to: identify directly related entities corresponding to the specific media event from the Internet media data; The database construction module is further used to supplement the industry knowledge graph database based on the specific media event and the corresponding directly related entity, wherein the specific media event is supplemented into the industry knowledge graph database as an abstract entity. 24. The apparatus according to claim 23, wherein the database construction module further extracts directly related entities corresponding to the specific media event by at least one of the following methods: Identify entities from text data based on entity recognition in natural language processing; identifying entities from image or video data based on image or video recognition processing; or Identify entities from audio or video data based on speech recognition processing. 25. The device according to any one of claims 21-23 is characterized in that the database construction module includes: a module for performing semantic disambiguation and entity linking on the extracted entities to identify the corresponding entities of the entities in the industry knowledge graph database. 26. The apparatus according to claim 25, wherein the module for performing semantic disambiguation and entity linking on the extracted entities further performs semantic disambiguation and entity linking in at least one of the following ways: Based on entity knowledge, semantic disambiguation and entity linking are performed independently for each extracted entity reference. Based on the topic consistency assumption, the associations of candidate entities in the knowledge base are utilized to perform consistent semantic disambiguation and entity linking on the extracted entity references. 27. The device according to claim 19, wherein the specific media event comprises a negative event, an emergency event, a crisis event, a mass event or a public opinion event. 28. A system for monitoring specific media events related to an industry, comprising: A data acquisition unit, configured to obtain industry data of one or more specific industries from a data source; a data processing unit, configured to process the industry data to extract entities related to the industry and corresponding entity attributes and/or entity relationships; A database construction unit, which performs semantic disambiguation and entity linking on the extracted entities, and is used to construct the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships; Database storage unit: used to store the constructed industry knowledge graph database; A media event monitoring unit is configured to acquire Internet media data, perform event detection, event evaluation, and screening based on the acquired Internet media data to acquire the specific media events related to the industry, and identify directly related entities corresponding to the specific media events, wherein the event detection comprises the following steps: performing topic classification on the Internet media data to acquire content for a specific topic; identifying entities involved in the acquired content; performing sentiment analysis on the acquired content and the identified entities, and filtering the acquired content based on the results of the sentiment analysis; and performing event discovery based on the filtered content to cluster media events and discover new media events. A database access unit is configured to access the industry knowledge graph database based on the directly related entities, and determine the indirect related entities corresponding to the specific media event based on the association relationships of the directly related entities in the industry knowledge graph database; A message sending unit is used to send a warning message to the directly related entity and/or the indirect related entity. 29. The system according to claim 28, wherein The data acquisition unit includes: a structured data acquisition unit, configured to obtain structured industry data from a third-party industry database, wherein the structured industry data includes a plurality of fields; The data processing unit includes: a structured data processing unit for performing data cleaning and extraction-transformation-loading (ETL) processing on the structured industry data before extracting entities related to the industry and corresponding entity attributes and/or entity relationships; The database construction unit includes: a database generation unit, which is used to generate the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 30. The system according to claim 28, wherein The data acquisition unit includes: an industry-related data acquisition unit, which is used to obtain industry-related data from an Internet data source using a web crawler technology, wherein the Internet data source includes an unstructured or semi-structured data source; The data processing unit includes: an industry-related data processing unit, configured to perform entity recognition and relationship extraction on the industry-related data using information extraction technology in natural language processing to extract the entities, entity attributes and/or entity relationships; The database construction unit includes: a database supplement/update unit, which is used to supplement or update the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 31. The system according to claim 28, wherein: The data acquisition unit includes: an industry-related data acquisition unit, configured to acquire industry-related data from an Internet data source in a query manner using an application programming interface (API), wherein the Internet data source includes an open data source; The data processing unit includes: an industry-related data processing unit, configured to perform data cleaning and Extract-Transform-Load (ETL) processing on the industry-related data before extracting entities and corresponding entity attributes and/or entity relationships related to the industry; The database construction unit includes: a database supplement/update unit, which is used to supplement or update the industry knowledge graph database based on the extracted entities, entity attributes and/or entity relationships. 32. The system according to claim 28, wherein: The data acquisition unit includes: a media data acquisition unit for acquiring industry-related Internet media data from an Internet data source using an application programming interface (API) or web crawler technology; The data processing unit includes: a media data processing unit for performing event detection, event evaluation, and screening on the Internet media data to extract specific media events related to the industry and identify corresponding directly related entities from the Internet media data; The database construction unit includes: a database supplement/update unit, which is used to supplement the industry knowledge graph database based on the specific media event and the corresponding directly related entity, wherein the specific media event is supplemented into the industry knowledge graph database as an abstract entity. 33. The system according to any one of claims 29-32 is characterized in that semantic disambiguation and entity linking are performed on the extracted entities to identify corresponding entities of the entities in the industry knowledge graph database. 34. The system according to claim 28, wherein the media event monitoring unit is further configured to: Topic classification of the content in the acquired Internet media data to obtain content targeting specific topics; identifying involved entities from the obtained content; Performing sentiment analysis on the obtained content and the identified entities, and filtering the obtained content based on the result of the sentiment analysis; Event discovery is performed based on the filtered content to cluster media events and discover new media events. 35. The system according to claim 34, wherein the media event monitoring unit is further configured to: The authenticity of the media events is analyzed based on their attributes, and the media events are sorted and/or filtered according to the analysis results. 36. The system according to claim 28, wherein the database access unit is further configured to: Based on the directly related entities, other entities and/or entity relationships in the industry knowledge graph database are queried under preset conditions to determine the non-directly related entities. 37. The system according to claim 28, wherein the database access unit is further configured to: Based on the directly related entities, data mining technology is used on other entities and/or entity relationships in the industry knowledge graph database to determine the non-directly related entities. 38. The system according to claim 28, wherein the specific media events include negative events, emergencies, crisis events, mass events or public opinion events.