JP2017504105A - System and method for in-memory database search - Google Patents

Abstract

Disclosed are systems and methods for identifying related entities using an entity co-occurrence knowledge base. In an embodiment, an entity identified in a search query is extracted using an entity's entity co-occurrence knowledge base extracted from an entity indexed corpus to present search results as related entities. An embodiment for generating search suggestions using fuzzy score matching with an entity co-occurrence knowledge base is also disclosed. Also, in an embodiment, a partial entity is extracted from the search query, a matching algorithm is executed based on the extracted entity type, and a search is performed on the entity co-occurrence knowledge base. Embodiments for generating search suggestions for related entities based on co-occurrence and / or fuzzy score matching are also disclosed. In these embodiments, the partial search questions are processed and suggested full question suggestions that are used as new search questions. Also disclosed is an embodiment for generating search suggestions using entity co-occurrence by extracting entities from a search query using an entity and trend co-occurrence knowledge base. Embodiments are also disclosed that enable geographic and named entity-based search capabilities in content management systems.
[Selection] Figure 1

Description

  The present invention relates generally to a method and system for information retrieval, and more particularly to a method for searching related entities using entity co-occurrence. The present invention relates generally to query enhancement, and more particularly to search suggestions using fuzzy score matching and entity co-occurrence in a knowledge base. The present invention relates generally to computer query processing, and more particularly to electronic search suggestions for related entities based on co-occurrence and / or fuzzy score matching. The present invention relates generally to a method and system for information retrieval, and more particularly to a method for obtaining search suggestions. The present invention relates generally to search engines and content management, and more particularly to extending search engine technology in content management systems to enable geotagging of digital content and enrichment of named entities.

  In the commercial context, a well-known search engine parses a set of search terms and returns a list of items (web pages in a typical search) that are categorized in some way. The best known solutions for performing searches are usually based on other users' historical references to build a search query database that is ultimately used to generate an index based on keywords. Yes. The user's search query includes one or more entities identified by names or attributes associated with the entities. Entities also include organizations, people, places, and / or times. In a typical search, when a user searches for information related to two specific organizations, the search engine returns an assortment result for a mixture of different entities with the same name or similar names. The latter solution results in the user finding a large number of documents that are not related to what the user is actually interested in.

  Accordingly, there is a need for a related entity search method that allows a user the ability to find related entities of interest.

  Users often use search engines to locate information of interest on either the Internet or any database system. A search engine typically operates by receiving a search query from a user and returning search results to the user. Search results are usually ordered by the search engine based on the relevance of each returned search result to the search query. Therefore, the quality of the search query is significantly important to the quality of the search results. However, search queries from users are only written incompletely or partially in most cases (eg, search queries do not contain enough words to generate a focused set of related results). Rather, it produces a large number of unrelated results), and sometimes there are also misspellings (eg Bill Smith is mistakenly spelled Bill Smith).

  One common solution for improving the quality of search results is to improve the search query. One way to improve search queries is by generating possible suggestions based on user input. For this reason, one solution proposes a method for identifying candidate question refinements for a given question from past questions submitted by one or more users. However, this solution is based on a question log that sometimes leads the user to uninteresting results. There are other solutions that use different technologies, but they are not accurate enough. Accordingly, there remains a need for a method for improving or enhancing search results from users to obtain more accurate results.

  Users often use search engines to locate information of interest on either the Internet or any database system. A search engine typically operates by receiving a search query from a user and returning search results to the user. Search results are usually ordered by the search engine based on the relevance of each returned search result to the search query. Therefore, the quality of the search query is significantly important to the quality of the search results. However, search queries from users are only written incompletely or partially in most cases (eg, search queries do not contain enough words to generate a focused set of related results). Rather, it produces a large number of unrelated results), and sometimes there are also misspellings (eg Bill Smith is mistakenly spelled Bill Smith).

  One common solution for improving the quality of search results is to improve the search query. One way to improve search queries is by generating possible suggestions based on user input. For this reason, one solution proposes a method for identifying candidate question refinements for a given question from past questions submitted by one or more users. However, this solution is based on a question log that sometimes leads the user to uninteresting results. There are other solutions that use different technologies, but they are not accurate enough. Accordingly, there remains a need for a method for improving or enhancing search results from users to obtain more accurate results and providing users with useful related entities that are of interest when the user types a search question. To do.

  The search engine has a number of features to provide predictions of user questions. Such predictions include automatic query completion and search suggestions. Today, such prediction methods are based on historical keyword references. Such historical references can be inaccurate because a keyword can refer to multiple topics in a single text.

  Further, the user search query includes one or more entities identified by names or attributes associated with the entities. These entities also include organizations, people, places, events, dates, and / or times. In a typical search, when a user searches for information related to two specific organizations, the search engine returns assortment results for a mixture of different entities with the same name or similar names. The latter solution results in the user finding a large number of documents that are not related to what the user is actually interested in.

  Accordingly, there is a need for a method for obtaining search suggestions more quickly and more accurately.

  Content management and document management systems for document versioning and collaborative object management are known. One non-limiting example is the Microsoft Sharepoint 2013® software and application toolset. Microsoft Sharepoint 2013 (R) is a family of software products developed by Microsoft for collaboration, file sharing and web publishing. This Sharepoint 2013 (registered trademark) gives a huge amount of content or information to the user, making it difficult for the user to find information most relevant to a particular situation. To alleviate these problems, Sharepoint 2013® provides a search engine to help users in finding the content they need. The user enters a keyword-based search query and the Sharepoint 2013® search engine is the most relevant found within the context of the Sharepoint 2013® platform when the content is syndicated. Send the result list back to the user.

  From time to time, users want to find content related to geographic entities in Sharepoint 2013 or other types of entities such as organizations or people referenced in documents. Sharepoint 2013 (registered trademark) does not provide a function for automatically extracting an entity from a document from a box. In particular, it does not support geotagging content for extracting geographic entities and resolving them relative to geographic locations. Sharepoint 2013 also does not support entity tagging to identify, disambiguate and extract named entities such as organizations or people in a document. However, Sharepoint 2013® search can be extended to allow for valid geographic and other entity-related searches, including entity-based search facets. Earlier versions of Sharepoint 2013® include a “FAST search” for Sharepoint, from which content processing pipelines can be extended through sandboxed applications, which is slow and The information that can be accessed is limited.

  Sharepoint 2013 (R) introduces a very open API that allows the addition of specialized linguistics such as concept extraction, relationship extraction, geotagging, summarization and sophisticated text analysis. Thus, there is an opportunity to extend the capabilities of the Sharepoint 2013® search engine to allow for geographic and other entity-based searches.

  A method for searching related entities using entity co-occurrence is disclosed. In one aspect of this disclosure, the method is used in a search system that includes a client / server type architecture. In some embodiments, the search system includes a user interface for a search engine that communicates with one or more server devices over a network connection. The server device includes an entity index corpus of electronic data, an entity co-occurrence knowledge base database, and an entity extraction computer module. The knowledge base is built as an in-memory database and includes other components such as one or more search controllers, multiple search nodes, collections of compressed data, and disambiguation modules. A search controller is selectively associated with one or more search nodes. Each search node can independently perform an ambiguous key search through the collection of compressed data and return a scored set of results to its associated search controller.

  In certain embodiments, a computer-implemented method receives a search query that includes one or more entities from a client computer by an entity extraction computer; each entity in the co-occurrence database is received by the entity extraction computer. One or more subsets of one or more entities from a search query by means of an entity extraction computer with the entity and one or more related entities in an electronic data corpus according to a co-occurrence database In response to the determination that each entity in the subset exceeds the reliability score of the co-occurrence database based on the accuracy of the index; ) To each of the entities in the plurality of extracted entities; the entity extraction computer saves an index ID for each of the plurality of extracted entities in the electronic data corpus, Indexed by an index ID corresponding to each of the two or more related entities; the search server computer positions the plurality of extracted entities, and at least two of the plurality of extracted entities co-occur Search the entity index type electronic data corpus to identify the index ID of the data record to be searched; and the search server computer has a data record corresponding to the identified index ID. To build that search results list; including the fact.

  In certain embodiments, the system comprises one or more server computers having one or more processors that execute computer readable instructions for a plurality of computer modules, which receive user input of search query parameters. A configured entity extraction module, wherein the entity extraction module further identifies each entity in the plurality of extracted entities with the accuracy of the co-occurrence of the extracted entity and one or more related entities in the electronic data corpus. Extract multiple entities from the search query parameters by comparing to an entity co-occurrence database that includes a confidence score representing and index each entity in the multiple extracted entities An identifier (index ID) is specified and an index ID for each of the plurality of extracted entities is saved in an electronic data corpus, the electronic data corpus being indexed by an index ID corresponding to each of the one or more related entities. An entity index to locate the plurality of extracted entities and to identify an index ID of a data record in which at least two of the plurality of extracted entities co-occur A search server module configured to search for an electronic data corpus of a type, the search server module further constructing a search result list having data records corresponding to the identified index ID Constructed.

  In another embodiment, a non-transitory computer readable medium receives user input of search query parameters by an entity extraction computer; the entity extraction computer extracts each entity in a plurality of extracted entities. Extracting a plurality of entities from a search query parameter by comparing to an entity co-occurrence database that includes a confidence score representing the likelihood of co-occurrence of the identified entities and one or more related entities in the electronic data corpus; Assigning an index identifier (index ID) to each entity in the plurality of extracted entities; each of the plurality of extracted entities by the entity extraction computer An index ID to be stored in an electronic data corpus, the electronic data corpus being indexed by an index ID corresponding to each of the one or more related entities; the search server computer extracts the plurality of extracted entities And searching the entity index type electronic data corpus to identify the index ID of the data record in which at least two of the plurality of extracted entities co-occur; and the search server computer identifies the A computer-executable instruction is stored including: building a search result list having data records corresponding to the index ID;

  A method for generating search suggestions by using fuzzy score matching and entity co-occurrence in a knowledge base is disclosed. In one aspect of this disclosure, the method is used in a search system that includes a client / server type architecture. In one embodiment, the search system includes a user interface for a search engine that communicates with one or more server devices over a network connection. The server device comprises an entity extraction computer module, an ambiguous score matching computer module, and an entity co-occurrence knowledge base database. The knowledge base is built as an in-memory database and also includes one or more search controllers, multiple search nodes, a collection of compressed data, and other hardware and / or software components such as a disambiguation computer module. ing. A search controller is selectively associated with one or more search nodes. Each search node can independently perform an ambiguous key search through the collection of compressed data and return a scored set of results to its associated search controller.

  In another aspect of this disclosure, the method identifies whether a search query refers to an entity and, if so, what type of entity to refer to, from a given search query, a partial entity It includes an entity extraction module that performs the extraction. The method further includes an ambiguous score matching module that spawns an algorithm based on the extracted entity type and performs a search against the entity co-occurrence knowledge base. Furthermore, portions of the question text that are not detected as corresponding to an entity are processed as conceptual features such as topics, facets, and key phrases that can be used to search the entity co-occurrence knowledge base. In one embodiment, the entity co-occurrence knowledge base includes a repository in which entities are indexed, among other things, as entity-to-entity, entity-to-topics, or entity-to-facet, which returns fast and accurate suggestions to the user. Helps complete search questions.

  In certain embodiments, a method is disclosed. The method receives user input of a search query parameter from a user interface by an entity extraction computer; the search query parameter is received by an entity extraction computer and an entity co-occurrence having co-occurrence instances of one or more entities in an electronic data corpus. Extracting one or more entities from the search query parameters by comparing to a database and identifying at least one entity type corresponding to the one or more entities in the search query parameters; Select an fuzzy matching algorithm to search the entity co-occurrence database to identify one or more records associated with the query parameter and its fuzzy match Grayed algorithm corresponds to at least one of the identified entities form. The method further includes searching the entity co-occurrence database by the fuzzy score matching computer using the selected fuzzy matching algorithm and one or more suggested searches from one or more records based on the search. Forming query parameters; and presenting one or more suggested search query parameters via a user interface by a fuzzy score matching computer.

  In another embodiment, a system is provided. The system includes one or more server computers having one or more processors that execute computer readable instructions for a plurality of computer modules, which are configured to receive user input of search query parameters from a user interface. The entity extraction module compares a search query parameter to an entity co-occurrence database having co-occurrence instances of one or more entities in the electronic data corpus and one in the search query parameter Further configured to extract one or more entities from the search query parameters by identifying at least one entity type corresponding to the above entities.The system further comprises an ambiguity score matching module configured to select an ambiguity matching algorithm that searches the entity co-occurrence database to identify one or more records associated with the search query parameter, the ambiguity matching A module corresponds to at least one identified entity type. The fuzzy score matching module further searches the entity co-occurrence database using the selected fuzzy matching algorithm and retrieves one or more suggested search query parameters from one or more records based on the search. Configured and configured to present one or more suggested search query parameters via a user interface.

  A method for generating related entity search suggestions based on co-occurrence and / or fuzzy score matching is disclosed. In one aspect of this disclosure, the method is used in a computer search system that includes a client / server architecture. In one embodiment, the search system includes a user interface for a search engine that communicates with one or more server devices over a network connection. The server device includes one or more processors that execute instructions for a plurality of special purpose computer modules including an entity extraction module and an ambiguous score matching module and an entity co-occurrence knowledge base database. The knowledge base is built as an in-memory database and includes other components such as one or more search controllers, multiple search nodes, a collection of compressed data, and a disambiguation module. A search controller is selectively associated with one or more search nodes. Each search node can independently perform an ambiguous key search through a collection of compressed data and return a scored set of results to its associated search controller.

  In another aspect of this disclosure, the method includes performing a partial entity extraction from a given search query by an entity extraction module to identify whether the search query refers to an entity, and if so, Including determining the format. In addition, the method includes generating an algorithm corresponding to the type of the extracted entity by the fuzzy score matching module and performing a search against the entity co-occurrence knowledge base. Furthermore, portions of the question text that are not detected as entities are processed as conceptual features such as topics, facets, and key phrases that can be used to search the entity co-occurrence knowledge base. An entity co-occurrence knowledge base where an entity already has a repository that is indexed as an entity-to-entity, entity-to-topic, or entity-to-facet, among other things, returns a quick and accurate suggestion to the user to complete the search query

  In yet another aspect of this disclosure, the completed search question is used as a new search question. The search system processes new search questions, performs entity extraction, finds the related entity with the highest score from the entity co-occurrence knowledge base, and presents the related entity in a drop-down list useful to the user .

  In certain embodiments, a method is disclosed. The method receives a user input of a partial search question parameter from a user interface by an entity extraction computer, the partial search question parameter having at least one incomplete search question parameter; Comparing the partial search query parameter with an entity co-occurrence database having co-occurrence instances of one or more first entities in the electronic data corpus, and at least one corresponding to the one or more first entities in the partial search query parameter One or more first entities are extracted from the partial search question parameters by identifying one entity type; and an ambiguous score matching computer Select ambiguous matching algorithm or search for entities co-occurrence database in order to identify the communication one or more records, the fuzzy matching algorithm, which corresponds to at least one of the identified entities form. The method further includes searching the entity co-occurrence database by the fuzzy score matching computer using the selected fuzzy matching algorithm and one or more first suggestions from one or more records based on the search. To present one or more first suggested search query parameters via a user interface by an ambiguous score matching computer; to form a completed search query parameter by an entity extraction computer Further receiving a user selection of one or more first suggested search query parameters; and extracting one or more second entities from the completed search query parameters by an entity extraction computer. The method further includes identifying, by an entity extraction computer, one or more entities associated with the one or more second entities to form one or more second suggested search query parameters. Searching the entity co-occurrence database; and presenting one or more second suggested search query parameters via the user interface by the entity extraction computer.

  In another embodiment, a system is disclosed. The system includes one or more server computers having one or more processors that execute computer-readable instructions for a plurality of computer modules, such that it receives user input of partial search query parameters from a user interface. Comprising a configured entity extraction module, the partial search question parameter having at least one unfinished search question parameter, the entity extraction module further including the partial search question parameter in the electronic data corpus as 1 Compared to an entity co-occurrence database having co-occurrence instances of one or more first entities and at least corresponding to one or more first entities in the partial search query parameters By identifying one entity type, configured to extract one or more first entity from part search query parameters. The system is further configured to select an ambiguity matching algorithm that searches the entity co-occurrence database to identify one or more records associated with the partial search query parameter, the ambiguity matching algorithm comprising: It corresponds to the identified entity type. The fuzzy score matching module further searches the entity co-occurrence database using the selected fuzzy matching algorithm and one or more first suggested searches from one or more records based on the search. A query parameter is formed and configured to present one or more first suggested search query parameters via a user interface. In addition, the entity extraction module further receives a user selection of one or more first suggested search question parameters to form a completed search query parameter, from which the one or more first search query parameters are received. An entity co-occurrence database to extract two entities and identify one or more entities associated with the one or more second entities to form one or more second suggested search query parameters It is configured to search and present one or more second suggested search query parameters via a user interface.

  A method for obtaining search suggestions associated with an entity using entity and feature co-occurrence is disclosed. In one aspect of this disclosure, the method is used in a search system that includes a client / server type architecture.

  Search systems that use methods that use entities stored on one or more servers allow entity databases and trend databases. Such database entities have a score to index based on a high score. A method for obtaining search suggestions combines information stored in both databases to generate a single list of search suggestions. The trend database provides previously searched questions from one or more users in the local network and / or the Internet. The entity database provides search suggestions based on entity extraction from multiple data available on the local network and / or the Internet. This list gives a more accurate and quick group of suggestions for the user.

  In certain embodiments, a computer-implemented method receives, by a computer, a search query that includes one or more data strings from a search engine, each entity corresponding to a subset of the one or more strings; Identifying one or more entities in one or more data strings based on comparing one or more entities against an entity database and a trend database; not identified by the computer as corresponding to at least one entity Identifying one or more features in one or more data strings; the computer assigns each of the one or more features to at least one of the one or more entities based on a matching algorithm; To specify an extraction score for each entity based on a score specified for each feature specified for each entity; by the computer, one or more having a score that is within a threshold distance from the extraction score for each entity Receiving a first search list including entities from the entity database; receiving from a trend database a second search list including one or more entities having a score that is within a threshold distance from each entity's extracted score; A computer generates a grand total list including a first search list and a second search list, and the entities of the grand total list are ranked according to the score of each total list; Providing a search suggested according; comprising.

  Disclosed herein are systems and methods that enable geographic entity-based searches in content management systems such as Microsoft Sharepoint 2013. The method described in the embodiment includes extending the Sharepoint 2013® search architecture by adding a geographic tagging web server. The system comprises a computer processor operatively associated with a computer memory and one or more I / O devices, wherein the processor and memory are adapted to operate one or more Sharepoint 2013® processes. Composed. The system also includes a computer memory and another computer processor operatively associated with the one or more I / O devices, wherein the processor and memory are adapted to host and provide processing for a geotagging web service. Configured. The Sharepoint 2013® system includes a crawling component, a content processing component, and a search index component to enable searching for content. The content processing component in Sharepoint 2013® search can extend its functionality by using Content Enrichment Web Server (CEWS) features.

  The method includes crawling content from different sources to obtain an array of crawl properties that are sent for content processing. During content processing, the trigger state determines whether the crawl property can benefit from additional processing to enrich the original content with additional geographic metadata properties. If the crawl property does not benefit from additional processing, the crawl property is mapped to a managed process and sent to the search index. If the crawl property benefits from external web service processing, the Simple Object Access Protocol (SOAP) to endpoints that CEWS can configure using Hypertext Transfer Protocol (HTTP) or other web service call methods Make a request. The entity enrichment service determines the type of content. If the content is in image format, its metadata, such as file location, is sent to an optical character recognition (OCR) engine, the original document is retrieved, processed asynchronously and converted to text, and Returned to the crawl component and recrawled in text format. If the content is in text format, the geotagging web service identifies the geographic metadata and associates it with the content as a managed property. The content is geotagged before being sent to the index component.

  Use the Sharepoint 2013® web portion or change the standard layout of the Sharepoint 2013® search with, among other things, standard web development tools such as HTML, HTML5, JavaScript, and CSS. Adds an additional search user interface (UI). The search UI may assist a user in performing a geographic search query or displaying geographic search results using, for example, but not limited to, a digital geographic feature such as a digital map. The search UI can also be enhanced to perform faceted searches using additional, enriched entities or associated metadata.

  Numerous other aspects, features, and benefits of this disclosure will become apparent from the following detailed description.

  The present disclosure can be better understood with reference to the following drawings. The components in the drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the present disclosure. In the figures, reference numerals indicate corresponding parts throughout the different views.

FIG. 2 is a block diagram illustrating an example environment of a computer system on which an embodiment of the invention operates. 4 is a flowchart illustrating a method for searching using entity co-occurrence according to one embodiment. FIG. 6 is a flow chart illustrating a simple search embodiment in which search results returned by the system include related entities of interest. 1 is a block diagram illustrating an example system environment in which an embodiment of the invention operates. FIG. 3 is a flowchart illustrating a method for providing search suggestions using fuzzy score matching and entity co-occurrence in a knowledge base according to one embodiment. FIG. 7 illustrates an example of a user interface that generates search suggestions using fuzzy matching and entity co-occurrence in the knowledge base of FIGS. 4-6. 1 is a block diagram illustrating an example system environment in which an embodiment of the invention operates. FIG. 6 is a flowchart illustrating a method for generating search suggestions for related entities based on co-occurrence and / or fuzzy score matching according to one embodiment. FIG. 9 is an exemplary embodiment of a user interface associated with the method shown in FIG. FIG. 5 is a block diagram illustrating a method for obtaining search suggestions based on an entity and trend database. FIG. 6 is a block diagram illustrating a method for obtaining search suggestions based on entity and trend databases by generating a list of suggestions based on individual scores of search suggestions in each database. FIG. 5 is a block diagram illustrating a method for obtaining search suggestions based on entity and trend databases by generating a list of suggestions based on the total search suggestion scores in both databases. It is a system architecture of content tagging and entity enrichment in a content management system. Fig. 4 illustrates the process of tagging and indexing contexts for named and geographic entity searches.

Definitions The following terms used herein have the following definitions.

  “Entity extraction” refers to an information processing method for extracting information such as name, location and organization.

  A “corpus” refers to a collection of one or more documents.

  “Features” is information derived at least in part from a document.

  “Event concept store” refers to a database of event template models.

  An “event” refers to one or more features characterized by at least real-time feature generation.

  An “event model” refers to a collection of data used to compare and identify a particular type of event.

  A “module” refers to a computer or software component suitable for performing at least one or more tasks.

  “Feature attribute” refers to metadata associated with the feature, for example, the location of the feature in the document, the confidence score, among others.

  “Fact” refers to an objective relationship between features.

  An “entity knowledge base” refers to a computer database containing features / entities.

  “Question” refers to a computer-generated request to retrieve information from one or more suitable databases.

  “Topics” refers to a set of semantic information derived at least in part from a corpus.

  “Geotagging” refers to the process of extracting geographic entities from an unstructured text file. Geotagging involves disambiguating entities to specific geographic locations and associated geographic metadata, such as geographic coordinates, geographic feature types, and other metadata.

  “Entity tagging” refers to the process of extracting named entities from unstructured text. Entity tagging includes entity disambiguation, entity name normalization, and accompanying entity metadata.

  “Named entity” refers to an individual, an organization, or a topic.

  A “geographic entity” refers to a geographical location or location.

  “Crawled properties” refers to content management system metadata that results from examining documents during crawling.

The preferred embodiment shown, respectively in the description accompanying drawings will be described in detail below. The above-described embodiments are merely examples. Those skilled in the art will recognize that the specific examples described herein may be replaced by numerous other components and embodiments within the scope of the present invention. Other embodiments may be used and / or other changes may be made without departing from the spirit or scope of the invention. The exemplary embodiments described in the detailed description are not meant to imply limitations on the subject matter presented herein.

  It will nevertheless be understood that no limitation of the scope of the invention is intended. Alternative and further embodiments of the features of the invention shown herein, as well as additional applications of the principles of the invention shown herein, which arise to those skilled in the art and possess the present disclosure, It should be considered within the scope of the invention.

  This disclosure describes systems and methods for detecting, extracting and validating events from multiple sources. Sources include sources that contain data relating to news sources, social media websites, and / or events.

  Various embodiments of the systems and methods disclosed herein collect data from different sources to identify independent events.

  FIG. 1 is a block diagram of a search system 100 according to the present invention. The search system 100 includes one or more client computing devices that include a processor that executes software modules associated with the search system 100, which includes a graphical user interface 102 that accesses the search engine 104 and over the network 108. A search query is communicated with the server device 106 in the form of binary data. In the exemplary embodiment, search system 100 is implemented in a client / server computing architecture. However, the search system 100 is implemented using other computer architectures (eg, stand-alone computers, mainframe systems with terminals, application service provider (ASP) models, peer-to-peer models, etc.). Also good. Network 108 comprises suitable hardware and software models that can communicate digital data between computing devices, such as a local area network, a wide area network, the Internet, a wireless network, a mobile telephone network, and the like. Thus, it will be apparent that the system 100 may be implemented over a single network 108 or using multiple networks 108.

  The user computing device 102 accesses a search engine 104 that includes a software model that can submit search queries. A search question is a parameter that is provided to the search engine 104 to indicate the information that it is desired to search. The search query is provided by the user or another software application in an appropriate data format (eg, integer, string, complex object) that is compatible with the search engine 104 parsing and processing routines. In some embodiments, the search engine 104 is a web-based tool that can be accessed through a browser or other software application on the user's computing device 102 and allows the user or software application to locate information on the World Wide Web. In one embodiment, search engine 104 is an application software module that is native to system 100 and that allows a user or application to locate information in the database of system 100.

  Server device 106, implemented as a single server device 106 or implemented in a distributed architecture across multiple server computers, includes an entity extraction module 110, an entity co-occurrence knowledge base 112, and an entity index corpus 114. ing. The entity extraction module 110 is a computer software and / or hardware module that can extract and disambiguate independent entities from a given set of questions, such as question strings, structured data, and the like. An entity is, for example, people, organization, geographical location, date and / or time. During extraction, one or more feature confirmation and extraction algorithms are used. Also, a score indicating the accuracy level of the feature that is correctly extracted together with the correct attribute is designated for each extracted feature. Taking into account the attributes of the features, the relative weight or relevance of each property is determined. In addition, a weighted score model is used to determine the degree of association between features.

  According to various embodiments, the entity co-occurrence knowledge base 112 is constructed as, but not limited to, an in-memory computer database (not shown) and includes one or more search controllers, multiple search nodes, compressed data. And other components (not shown) such as a disambiguation computer module. A search controller is selectively associated with one or more search nodes. Each search node can independently perform an ambiguous key search through a collection of compressed data and return a set of scored results to its associated search controller.

  The entity co-occurrence knowledge base 112 includes related entities that are feature-based and ranked by a reliability score. A method of linking features, essentially using a weighted model to determine which entity type is most important, which has a greater weight, and based on the confidence score, Various methods are used such as determining how reliable the extraction of the correct features was done. The entity indexed corpus 114 includes data from multiple sources, such as the Internet, having a large or live corpus.

  FIG. 2 is a flowchart illustrating a method 200 for searching related entities using entity co-occurrence implemented in the search system 100 as shown in FIG. According to various embodiments, prior to initiating method 200, an entity indexed corpus 114 similar to that shown in FIG. 1 may include a plurality of sources such as a bulk corpus of electronic data or a live corpus (eg, Data from the Internet, websites, blogs, word processing files, plain text files). The entity indexed corpus 114 includes a plurality of indexed entities that are constantly updated as new data is discovered.

  In one embodiment, the method 200 begins at step 202 when a user or software application of the computing device 102 provides the search engine 104 with one or more search questions that include one or more entities. The search questions given in step 202 are processed by the search system 100 from 1 to n each time. The search query in step 202 is a combination of keywords such as, for example, a string, structured data, or other suitable data format. In the exemplary embodiment of FIG. 2, the search query keywords are entities representing people, organizations, geographic locations, dates and / or times.

  The search query from step 202 is then processed for entity extraction in step 204. In this step, the entity extraction module 110 processes the search queries from step 202 as entities and compares them all against the entity co-occurrence knowledge base 112 to extract as many entities as possible and disambiguation. To do. During extraction, one or more feature confirmation and extraction algorithms are used. In addition, a score indicating the accuracy level of a feature that is correctly extracted with a correct attribute is designated for each extracted feature. Considering feature attributes, the relative weight or relevance of each feature is determined. In addition, a weighted score model is used to determine the degree of association between features.

  Further, a method for linking features, essentially using a weighted model to determine which entity type is most important, which has the greater weight, and to the reliability score Based on this, various methods are used, such as determining how reliably the correct feature extraction has been performed. Once the entities are extracted and ranked based on the confidence score, an index ID, which in some cases is a number, is assigned to the extracted entities in step 206.

  Next, in step 208, a search based on the entity index ID specified in step 206 is performed. In the search step 208, the extracted entities are positioned in the entity index corpus 114 using standard indexing methods. Once the extracted entity is located, it continues to the entity association step 210. In the entity association step 210, all data such as documents, videos, pictures, files, etc. where at least two extracted entities overlap is extracted from the entity index corpus 114. Finally, in step 212, a list of potential results is constructed, sorted by relevance, and presented to the user as search results. The resulting list then shows only the links to the data where the user finds relevant entities of interest.

  FIG. 3 is a specific example of a method 300 for searching related entities using entity co-occurrence, as described above in connection with FIG. As described with respect to FIG. 2, according to various embodiments, prior to the start of method 300, an entity-indexed corpus 114 similar to that described in FIG. The data from the source (Internet) is supplied. The entity indexed corpus 114 includes a plurality of indexed entities that are constantly updated as new data is discovered.

  In this exemplary embodiment, the user searches for information about “Jobs” of the company “Apple”. Thus, the user enters one or more entities (eg, the search query at step 302) through the user interface 102, and the user interface involves, but is not limited to, a search engine 104 as described with respect to FIG. Interface. By way of example and not limitation, the user enters a combination of entities such as “Apple + Jobs”. The search engine 104 then generates search questions at step 302 and sends the queries to the server device 106 for processing. In the server device 106, the entity extraction module 110 performs the entity extraction of step 304 from the search query input of step 302.

  The entity extraction module 110 then processes the search questions entered in step 302, eg, “Apple” and “Jobs”, as entities, and compares them all against the entity co-occurrence knowledge base 112, as many as possible. Extract and disambiguate entities. During extraction, one or more feature confirmation and extraction algorithms are used. Also, a score indicating the accuracy level of the feature that is correctly extracted with the correct attribute is designated for each extracted feature. Considering feature attributes, the relative weight or relevance of each feature is determined. In addition, a weighted score model is used to determine the degree of association between features.

  In addition, a method of linking features, essentially using a weighted model to determine which entity type is most important, which has the greater weight, and the reliability score Based on this, various methods are used such as determining how reliable the extraction of the correct features has been done. As a result, a table 306 containing entities and co-occurrence is generated. The table 306 shows the entity “apple” and its co-occurrence, in this case Apple and Jobs, Apple and Steve Jobs. Table 306 also includes Apple and Organization A, which is known to be relevant because Organization A does business with Apple and generates “jobs” in Organization A. Other co-occurrences are found with low importance. Thus, Apple and Jobs have the highest score (1) and are therefore listed at the top, then Apple and Steve Jobs have the second highest score (0.8), and last In addition, Apple and other Organizations A are listed at the bottom with the lowest score (0.3).

  Once the entities are extracted and ranked based on the confidence score, an index ID, which may be a number in some cases, is assigned to the extracted entities in step 308. The table 310 shows index IDs specified for the extracted entities. Thus, the table 310 indicates “Apple” with index ID1, “Jobs” with index ID2, “Steve Jobs” with index ID3, and “Organization” with index ID4.

  A search step 312 based on the entity index ID (308) is then performed. In search step 312, extracted entities such as “Apple”, “Jobs”, “Steve Jobs”, and “Organization A” are positioned in the entity indexed corpus 114 using standard indexing methods.

  After positioning the extracted entity in the entity index corpus 114, the entity association step 314 is followed. In the entity association step 314, all data such as documents, videos, pictures, files, etc. on which at least two extracted entities are superimposed are extracted from the entity index corpus 114 and a list of links is constructed as a search result. (Step 318). By way of example and not limitation, the table 316 shows how the extracted entities are associated with the data in the entity index corpus 114. In table 316, documents 1, 4, 5, 7, 8, and 10 show a superposition of two extracted entities, so links for those documents are shown as search results in step 318.

  FIG. 4 is a block diagram of a search computer system 400 according to the present invention. Search system 400 includes one or more user interfaces 402 to search engine 404 that communicate with server device 406 over network 408. In this embodiment, the search system 400 is implemented in one or more special purpose computers and computer modules described below, including through a client / server type architecture. However, the search system 400 may be implemented using other computer architectures (eg, stand-alone computers, mainframe systems with terminals, ASP models, peer-to-peer models, etc.). In one embodiment, search computer system 400 includes multiple networks, such as a local area network, a wide area network, the Internet, a wireless network, a mobile telephone network, and the like.

  Search engine 404 includes a user interface such as a web-based tool that allows a user to locate information on the World Wide Web. The search engine 404 also includes user interface tools that allow a user to locate information within an internal database system. Server device 406, implemented in a single server device 406 or in a distributed architecture across multiple server computers, includes an entity extraction module 410, an ambiguity score matching module 412, and an entity co-occurrence knowledge base database 414. including.

  The entity extraction module 410 is a hardware and / or software module configured to extract and disambiguate independent entities from a given set of questions such as question strings, partial questions, structured data, etc. . An entity is, for example, people, organization, geographical location, date and / or time. During extraction, one or more feature confirmation and extraction algorithms are used. Also, a score indicating the accuracy level of the feature that is correctly extracted together with the correct attribute is designated for each extracted feature. Taking into account the attributes of the features, the relative weight or relevance of each property is determined. In addition, a weighted score model is used to determine the degree of association between features.

  The ambiguity score matching module 412 includes a plurality of algorithms that are selected according to the type of entity extracted from a given search query. The function of the algorithm is to determine whether a given search query received via user input and other searched strings identified by the algorithm are similar to each other or approximately match a given pattern string It is. Fuzzy matching is also known as fuzzy string matching, inexact matching, and approximate matching. The entity extraction module 410 and the fuzzy score matching module 412 work in conjunction with the entity co-occurrence knowledge base 414 to generate search suggestions for the user.

  According to various embodiments, the entity co-occurrence knowledge base 414 is constructed as, but not limited to, an in-memory computer database and includes one or more search controllers, multiple search nodes, a collection of compressed data, and Includes components such as disambiguation modules. A search controller is selectively associated with one or more search nodes. Each search node can independently perform an ambiguous key search through a collection of compressed data and return a set of scored results to its associated search controller.

  The entity co-occurrence knowledge base 414 includes related entities that are feature-based and ranked by a reliability score. A method of linking features, essentially using a weighted model to determine which entity type is most important, which has a greater weight, and based on the confidence score, Various methods are used such as determining how reliable the extraction of the correct features was done.

  FIG. 5 is a flowchart illustrating a method 500 for generating search suggestions using fuzzy score matching and entity co-occurrence in a knowledge base. The method 500 is implemented in a search system 400 similar to that shown in FIG.

  In some embodiments, the method 500 begins at step 502 when a user begins typing a search query into the search engine interface 402 shown in FIG. When a search query is typed at step 502, the search system 400 performs an on-the-fly process. According to various embodiments, the search query input of step 502 is complete or partial and is a correct spelling or misspelling. Thereafter, in the search system 400, the partial entity extraction step 504 from the search query input in step 502 is performed. The partial entity extraction step 504 performs a quick search on the entity co-occurrence knowledge base 414 to identify whether the search query entered in step 502 is an entity, and if so, what type of entity Is identified. According to various embodiments, the search query input of step 402 refers to, among others, an individual, an organization, a location or location, and a date. Once the entity type of the search query entry is identified, the fuzzy score matching module 412 selects a corresponding fuzzy matching algorithm at step 506. For example, if a search query is identified as an entity pointing to an individual, the fuzzy score matching module 412 extracts different components of the individual's name including, for example, first name, middle name, last name, and title. To select a string matching algorithm for the individual. In another embodiment, if the search question is identified as an entity pointing to an organization, the fuzzy score matching module 412 may use a string matching algorithm for the organization that includes the identifying term such as school, university, company, company, etc. select. The fuzzy score matching module 412 then selects a string matching algorithm corresponding to the identified entity type in the search query input for superior search. Once the string matching algorithm is adjusted for that identified entity type, an ambiguous score matching step 508 is performed.

  In the fuzzy score matching step 508, the extracted entity (s) and non-entities are selected and compared against the entity co-occurrence knowledge base 414. The extracted entity (s) is an incomplete name of the individual, for example, the first letter of the first name and last name, an organization abbreviation, for example, “UN” meaning “United Nation”. , Abbreviations, and nicknames. The entity co-occurrence knowledge base 414 has already registered multiple records that are indexed as structured data, such as entity-to-entity, entity-to-topics, and entity-to-facts, among others. The latter allows the fuzzy score matching of step 508 to be performed very quickly. The ambiguous score matching of step 508 uses common string metrics such as, but not limited to, Levenstein distance, strcmp95, ITF scoring, etc. The Levenstein distance between two words refers to the minimum number of single character edits required to change one word to another.

  Finally, when the fuzzy score matching step 508 finishes comparing and searching the search query for all records in the entity co-occurrence knowledge base 414, is it best matching the given pattern string (ie, the search query input of step 502)? Or the record closest to the match is selected as the first candidate for search suggestions in step 510. Other records that are not very close to matching a given pattern string are placed in decreasing order under the first candidate. The search suggestion at step 510 is presented to the user in a drop-down list of possible matches, which the user may or may not ignore.

  FIG. 6 illustrates an example user interface 600 based on a method for generating search suggestions using the fuzzy score matching and entity co-occurrence knowledge bases described with respect to FIGS. 4-5. In this example, the user enters a partial query 604 in search box 606 through a search engine interface similar to that shown in FIG. By way of example and not limitation, the partial query 604 is an incomplete name of an individual such as “Michael J” as shown in FIG. This is considered partial question 604 because the user has not yet selected search button 608 or otherwise submitted partial query 604 to search system 400 to perform the actual search and have not obtained any results. It is done.

  Following the method 500 (FIG. 5), when the user types “Michael J”, the entity extraction module 410 performs a quick search of the first word (Michael) on the fly against the entity co-occurrence knowledge base 414. The entity type, in this example the entity refers to the name of the person. As a result, the fuzzy score matching module 412 selects a string matching algorithm tailored to the person's name. Personal names are written in different forms, for example using only the initials (short form), or the first letters of the first and last names, or the first and middle names, the initial and last names, or a combination thereof. Found in other databases. The ambiguous score matching module 412 uses a common string metric such as Levenstein distance to determine and specify a score for an entity, topic, or fact in the entity co-occurrence knowledge base 414 that matches the entity “Michael”. . In this example, Michael matches a huge amount of records with that name. However, when the user types the next letter “J”, the fuzzy score matching module 412 performs another comparison based on the Levenstein distance for all co-occurrence with Michael in the entity co-occurrence knowledge base 414. Carry out. The entity co-occurrence knowledge base 414 then selects all possible matches with the highest score for “Michael J”. For example, the ambiguous score matching module 412 returns search suggestions 610 such as “Michael Jackson”, “Michael Jordan”, “Michael J. Fox”, or in some cases “Michael Dell” to the user. The user can then select one of the suggested individuals from the drop-down list to complete the search question. Extending the above example, a question like “Michael the basketball player” is an entity co-ordination for “Michael” in personal entity name change and “the basketball player” in co-occurrence features like key phrases, facts and topics. Based on the results returned by searching the knowledge base, it leads to the suggestion of “Michael Jordan”. As another example, “Alexander the actor” invites the suggestion of “Alexander Polinsky”. It will be apparent to those skilled in the art that existing platforms cannot generate suggestions as described above.

  FIG. 7 is a block diagram of a search system 700 according to the present invention. The search system 700 includes one or more user interfaces 702 for a search engine 704 that communicates with a server device 706 via a network 708. In this embodiment, the search system 700 is implemented in a client / server architecture, although the search system 700 may be implemented in other computer architectures (eg, stand-alone computers, mainframe systems with terminals, ASP models, peers). 2), and multiple networks such as a local area network, a wide area network, the Internet, a wireless network, a mobile phone network, etc.

  Search engine 704 includes, but is not limited to, an interface via a web-based tool that allows a user to locate information on the World Wide Web. Search engine 704 also includes tools that allow a user to locate information within an internal database system. Server device 706, implemented in a single server device 706 or in a distributed architecture across multiple server computers, includes an entity extraction module 710, an ambiguity score matching module 712, and an entity co-occurrence knowledge base database 714. including.

  The entity extraction module 710 is a hardware and / or software module that can extract and disambiguate independent entities from a given set of questions such as question strings, partial questions, structured data, and the like. An entity is, for example, people, organization, geographical location, date and / or time. During extraction, one or more feature confirmation and extraction algorithms are used. Also, a score indicating the accuracy level of the feature that is correctly extracted together with the correct attribute is designated for each extracted feature. Considering feature attributes, the relative weight or relevance of each feature is determined. In addition, a weighted score model is used to determine the degree of association between features.

  The fuzzy score matching module 712 includes a plurality of algorithms that are adjusted or selected according to the type of entity extracted from a given search question. The function of the algorithm is to determine whether a given search query (input) and the searched and suggested strings are similar to each other or approximately match a given pattern string. Fuzzy matching is also known as fuzzy string matching, inexact matching, and approximate matching. The entity extraction module 710 and the fuzzy score matching module 712 work in conjunction with the entity co-occurrence knowledge base 714 to generate search suggestions for the user.

  According to various embodiments, the entity co-occurrence knowledge base 714 is constructed as, but not limited to, an in-memory computer database and includes one or more search controllers, a plurality of search nodes, a collection of compressed data, and Includes components such as disambiguation modules. A search controller is selectively associated with one or more search nodes. Each search node can independently perform an ambiguous key search through a collection of compressed data and return a set of scored results to its associated search controller.

  The entity co-occurrence knowledge base 714 includes related entities that are feature-based and ranked by a reliability score. A method of linking features, essentially using a weighted model to determine which entity type is most important, which has a greater weight, and based on the confidence score, Various methods are used such as determining how reliable the extraction of the correct features was done.

  FIG. 8 is a flowchart illustrating one embodiment of a method 800 for generating related entity search suggestions based on co-occurrence and / or fuzzy score matching. The method 800 is implemented in a search system 700 similar to that described with respect to FIG.

  In some embodiments, the method 800 begins when the user types a search query at step 802 in the search engine 704 described above with respect to FIG. When a search query is typed, the search system 700 performs an on-the-fly process. According to various embodiments, the search query is complete and / or partial, correct spelling and / or misspelled. A search query partial entity extraction step 804 is then performed. The partial entity extraction step 804 performs a quick search against the entity co-occurrence knowledge base 714 to identify whether the search question includes entities, and if so, identifies the type of entity. According to various embodiments, search query entities refer to individuals, organizations, locations or places, and dates, among others. If there is an entity type, the fuzzy score matching module 712 selects a corresponding fuzzy matching algorithm at step 806. For example, if a search query is identified as an entity pointing to an individual, the fuzzy score matching module 712 may be for individuals who can extract different components of the individual's name, including first name, middle name, last name, and title. Adjust or select a string matching algorithm. In another embodiment, if the search query is identified as an entity that points to an organization, the fuzzy score matching module 712 may use a string matching algorithm for the organization that includes the identifying term such as school, university, company, company, etc. Adjust or select. Therefore, the fuzzy score matching module 712 adjusts or selects a string matching algorithm for the type of entity to facilitate searching. Once the string matching algorithm is adjusted or selected to correspond to the type of entity, an ambiguous score matching step is performed at step 808.

  In the fuzzy score matching step 808, the extracted entity (s) and non-entities are selected and compared against the entity co-occurrence knowledge base 714. The extracted entity (s) is an incomplete name of the individual, for example, the first letter of the first name and last name, an organization abbreviation, for example, “UN” meaning “United Nation”. , Abbreviations, and nicknames. The entity co-occurrence knowledge base 714 has already registered a plurality of records that are indexed as structured data, such as entity-to-entity, entity-to-topics, and entity-to-facts, among others. This allows the fuzzy score matching of step 808 to be done quickly. Ambiguous score matching uses, but is not limited to, common string metrics such as Levenstein distance, strcmp95, ITF scoring. The Levenstein distance between two words refers to the minimum number of single character edits required to change one word to another.

  When the ambiguity score matching of step 808 finishes comparing and searching the search query against all records in the entity co-occurrence knowledge base 714, the record that best matches or is closest to the given pattern string in the search query input is In step 810, it is selected as the first candidate for search suggestions. Other records that are not very close to matching the search query input with a given pattern string are placed in decreasing order under the first candidate. The search suggestions at step 810 are presented to the user in a drop-down list of possible matches that the user selects to complete the question.

  In another embodiment, after the user selects a match of interest, search system 700 takes the selection as a new search question at step 812. Thereafter, an entity extraction step 814 from the new search query is performed. During extraction, one or more feature confirmation and extraction algorithms are used. In addition, a score indicating the accuracy level of a feature that is correctly extracted with a correct attribute is designated for each extracted feature. Considering feature attributes, the relative weight or relevance of each feature is determined. In addition, a weighted score model is used to determine the degree of association between features. The entity extraction module 710 then performs a search against the entity co-occurrence knowledge base 714 to find related entities based on the co-occurrence with the highest score (step 816). Finally, at step 818, a search suggestion drop-down list containing related entities is presented to the user prior to performing the actual data search in the electronic document corpus.

  FIG. 9 is an exemplary embodiment of a user interface 900 associated with a method 800 for generating related entity search suggestions based on co-occurrence and / or fuzzy score matching. In this example, the user enters a partial query 904 into search box 906 through a search engine interface 902 similar to that shown in FIG. By way of example and not limitation, the partial question 304 is an incomplete name of an individual such as “Michael J” as shown in FIG. This is considered partial query 904 because the user has not yet selected search button 908 or otherwise submitted partial query 904 to search system 100 to perform the actual search and have not obtained any results. It is done.

  Following the method 800, when the user types “Michael J”, the entity extraction module 710 performs a quick search of the first word (Michael) on the entity co-occurrence knowledge base 714 on the fly to In this example, an entity refers to a person's name. As a result, the fuzzy score matching module 712 selects a string matching algorithm tailored to the person's name. Personal names are written in different forms, for example using only the initials (short form), or the first letters of the first and last names, or the first and middle names, the initial and last names, or a combination thereof. Found in other databases. The ambiguous score matching module 712 uses a common string metric such as Levenstein distance to determine and specify scores for entities, topics or facts in the entity co-occurrence knowledge base 714 that matches the entity “Michael”. . In this example, Michael matches a huge amount of records with that name. However, when the user types the next letter “J”, the fuzzy score matching module 712 performs another comparison based on the Levenshtein distance for all co-occurrence with Michael in the entity co-occurrence knowledge base 714. Carry out. The entity co-occurrence knowledge base 714 then selects all possible matches with the highest score for “Michael J”. For example, the ambiguous score matching module 712 returns search suggestions 910 such as “Michael Jackson”, “Michael Jordan”, “Michael J. Fox”, or in some cases “Michael Dell” to the user. The user can then select one of the suggested individuals from the drop-down list, or ignore the suggestion and continue typing. Extending the above example, questions such as “Michael the basketball player” will ask for “Michael” in personal entity name changes and “the basketball player” in co-occurrence features such as key phrases, facts, topics, etc. Based on the results returned by searching the knowledge base, it leads to the suggestion of “Michael Jordan”. As another example, “Alexander the actor” invites the suggestion of “Alexander Polinsky”. As will be apparent to those skilled in the art, existing search platforms cannot provide the suggestions generated as described above.

  In this embodiment, the user selects “Michael Jordan” from the drop-down list to complete the partial question 904, as shown in FIG. That selection is then processed by search system 700 as a new search query 912. Thereafter, entity extraction from the new search query 912 is performed. During extraction, one or more feature confirmation and extraction algorithms are used. In addition, a score indicating the accuracy level of a feature that is correctly extracted with a correct attribute is designated for each extracted feature. Considering feature attributes, the relative weight or relevance of each feature is determined. In addition, a weighted score model is used to determine the degree of association between features. The entity extraction module 710 then performs a search for “Michael Jordan” against the entity co-occurrence knowledge base 714 to find related entities based on the co-occurrence with the highest score. Finally, a drop-down list of search suggestions 914 containing related entities is presented to the user prior to performing the actual data search by clicking the search button 908. The systems and methods described with respect to FIGS. 7-9 are quick and convenient for the user because the user can find useful relationships.

  FIG. 10 is a block diagram of a search system 1000 according to the present invention. The search system 1000 includes a search engine 1002, which includes one or more user interfaces that allow data input from a user, eg, user questions.

  The search system 1000 includes one or more databases. Such databases include an entity database 1004 and a trend database 1006. The database is stored on a local server or a web-based server. Thus, the search system 1000 is implemented in a client / server architecture, but the search system 1000 may be implemented in other computer architectures, such as a stand-alone computer, a mainframe system with terminals, an ASP model, a peer-to-peer, etc. It may be implemented using a peer model, etc., as well as multiple networks such as a local area network, a wide area network, the Internet, a wireless network, a mobile telephone network, and the like.

  Search engine 1002 includes, but is not limited to, web-based tools that allow users to locate information on the World Wide Web. The search engine 1002 also includes tools that allow the user to locate information within the internal database system.

  The entity database 1004 may be implemented as a single server or in a distributed architecture across multiple servers. The entity database 1004 allows a set of entity questions such as question strings, structured data, and the like. Such a set of entity queries is extracted in advance from a plurality of corpora available on the Internet and / or local network. Entity questions are indexed and scored. Entities include, for example, people, organizations, geographic locations, dates and / or times. During extraction, one or more feature confirmation and extraction algorithms are used. Also, a score indicating the accuracy level of the feature that is correctly extracted together with the correct attribute is designated for each extracted feature. Taking into account the attributes of the features, the relative weight or relevance of each property is determined. In addition, a weighted score model is used to determine the degree of association between features.

  Trend database 1006 may be implemented as a single server or in a distributed architecture across multiple servers. The trend database 1006 allows a set of entity questions such as question strings, structured data, etc. Such a set of entity questions is extracted in advance from historical questions performed by the user and / or multiple users in the Internet and / or local network. Entity questions are indexed and scored. Entities include, for example, people, organizations, geographic locations, dates and / or times. During extraction, one or more feature confirmation and extraction algorithms are used. Also, a score indicating the accuracy level of the feature that is correctly extracted together with the correct attribute is designated for each extracted feature. Taking into account the attributes of the features, the relative weight or relevance of each property is determined. In addition, a weighted score model is used to determine the degree of association between features.

  The entity database 1004 and the trend database 1006 comprise an entity co-occurrence knowledge base, which is constructed as an in-memory database (not shown), but is not limited to one or more search controllers, Includes other components (not shown) such as search nodes, collections of compressed data, and disambiguation modules. A search controller is selectively associated with one or more search nodes. Each search node can independently perform an ambiguous key search through the collection of compressed data and return a scored set of results to its associated search controller.

  The co-occurrence knowledge base includes related entities that are feature-based and ranked by a reliability score. A method of linking features, essentially using a weighted model to determine which entity type is most important, which has a greater weight, and based on the confidence score, Various methods are used such as determining how reliable the extraction of the correct features was done.

  The search system 1000 compares user queries in the search engine 1002 against the entity database 1004 and the trend database 1006. The auto-completion mode in the search engine 1002 is enabled from both databases: the entity database 1004 and the trend database 1006. Search system 1000 expands a list of search suggestions 1008 to the user, and such a list is generated and indexed based on the ambiguity score specified for each entity suggestion in the database. The score for each entity suggestion is automatically specified by the search system 1000 and / or manually by the system supervisor. Entity suggestions are ordered from highest to lowest relevance based on the score obtained by each entity. In addition, the scores in the trend database 1006 are specified using trends and question frequencies from one or more users in the local network and / or the Internet.

  The entity suggestions in each database are compared among them, then indexed and ordered by the rank obtained in the score, thus searching suggestions 1008 that synthesize entity suggestions in both databases, ie, entity database 1004 and trend database 1006. Is shown to the user. If the user selects an suggestion from the list or selects another result from the suggestion list, the search system 1000 saves such information in the trend database 1006. Accordingly, a self-learning system that increases the reliability and accuracy of the search system 1000 is allowed. In summary, the trend co-occurrence knowledge base is continuously updated with features extracted from user questions and selected suggestions to provide a means for on-the-fly learning, which improves search relevance and accuracy. To do. Furthermore, the trend co-occurrence knowledge base can be populated by different users using the system and by automatic methods such as trend detection modules.

  FIG. 11 is a block diagram of a search system 1100 according to the present invention. The search system 1100 includes a search engine 1102, which includes one or more user interfaces that allow data input from the user, such as user questions.

  The search system 1100 includes one or more databases. Such databases include an entity database 1104 and a trend database 1106. The database is stored on a local server or a web-based server. Thus, the search system 1100 is implemented in a client / server architecture, although the search system 1100 can be implemented in other computer architectures, such as a stand-alone computer, a mainframe system with terminals, an ASP model, peer-to-peer, etc. It may be implemented using a peer model, etc., as well as multiple networks such as a local area network, a wide area network, the Internet, a wireless network, a mobile telephone network, and the like.

  In some embodiments, the search system 1100 starts when a user enters one or more entities (in a search query) through the search engine 1102 user interface. The search question is, for example, a combination of keywords in a string data format, structured data, etc. These keywords are entities that represent people, organizations, geographic locations, dates and / or times. In this embodiment, “Indiana Na” is used as the search question.

  “Indiana Na” is then processed for entity extraction. The entity extraction model treats search queries such as “Indiana Na” as entities and compares them all against the entity co-occurrence knowledge base in the entity database 1104 and the trend database 1106, as many entities as possible. Are extracted and disambiguated. In addition, portions of the question text that are not detected as entities (eg, individuals, organizations, locations) can be used to search conceptual features (eg, topics, facts, Key phrase). During extraction, one or more feature confirmation and extraction algorithms are used. Also, a score indicating the accuracy level of the feature that is correctly extracted together with the correct attribute is designated for each extracted feature. Taking into account the attributes of the features, the relative weight or relevance of each property is determined. In addition, a weighted score model is used to determine the degree of association between features.

  In this embodiment, the entity database 1104 shows a list of search suggestions as a list 1108 of indexed and ranked entity suggestions. The trend database 1106 shows a list of search suggestions as a trend-based suggestion list 1110 that is indexed and ranked. The search system 1100 then builds a search suggestion list 1112 based on what is provided by the entity database 1104 and the trend database 1106. This search suggestion list 1112 is indexed and ranked based on the individual scores of each entity suggestion in each database, so the highest relevance is shown first, followed by the low relevance results.

  In search system 1100, an exemplary use for obtaining search suggestions is disclosed. Search suggestion list 1112 shows suggestions based on the “Indiana Na” user question. As a result, “Indiana Name” appears first for that entity based on the individual score 0.9, then “Indiana Nasca” is shown as the result of the individual score 0.8, and finally the individual score Based on 0.7, “Indiana Nashville” is indicated. Individual scores are compared using a list of entity suggestions 1108 and a trend-based suggestion list 1110 without applying possible recurring entities.

  FIG. 12 is a block diagram of a search system 1200 according to the present invention. The search system 1200 includes a search engine 1202, which includes one or more user interfaces that allow data input from the user, such as user questions.

  Search system 1200 includes one or more databases. Such databases include an entity database 1204 and a trend database 1206. The database is stored on a local server or a web-based server. Thus, while the search system 1200 is implemented in a client / server architecture, the search system 1200 may be implemented in other computer architectures such as stand-alone computers, mainframe systems with terminals, ASP models, peer-to-peer, etc. It may be implemented using a peer model, etc., as well as multiple networks such as a local area network, a wide area network, the Internet, a wireless network, a mobile telephone network, and the like.

  In some embodiments, the search system 1200 starts when a user enters one or more entities (in a search query) through the search engine 1202 user interface. A search question is a combination of keywords in a string, structured data, etc., for example. These keywords are entities that represent people, organizations, geographic locations, dates and / or times. In this embodiment, “Indiana Na” is used as the search question.

  “Indiana Na” is then processed for entity extraction. The entity extraction model treats search queries such as “Indiana Na” as entities and compares them all against the entity co-occurrence knowledge base in the entity database 1204 and trend database 1206, as many entities as possible. Are extracted and disambiguated. In addition, portions of the question text that are not detected as entities (eg, individuals, organizations, locations) can be used to search conceptual features (eg, topics, facts, Key phrase). During extraction, one or more feature confirmation and extraction algorithms are used. Also, a score indicating the accuracy level of the feature that is correctly extracted together with the correct attribute is designated for each extracted feature. Taking into account the attributes of the features, the relative weight or relevance of each property is determined. In addition, a weighted score model is used to determine the degree of association between features.

  In this embodiment, the entity database 1204 shows a list of search suggestions as a list 1208 of pre-indexed and ranked entity suggestions. Similarly, the trend database 1206 shows a list of search suggestions as a trend-based suggestion list 1210 that is pre-indexed and ranked. The search system 1200 then builds a search suggestion list 1212 based on what is provided by the entity database 1204 and the trend database 1206. This search suggestion list 1212 is indexed and ranked based on the overall score of each entity suggestion in both databases, so the highest relevance is shown first, followed by the low relevance results.

  In search system 1200, an exemplary use for obtaining search suggestions is disclosed. Search suggestion list 1212 shows suggestions based on the “Indiana Na” user question. As a result, “Indiana Nasca” appears first based on an overall score of 1.4 resulting from the sum of the score 0.8 in the entity suggestion list 1208 and the score 0.6 in the trend-based suggestion list 1210. Similarly, “Indiana Name” is shown as a result of an overall score of 0.9, and finally “Indiana Nashville” is shown based on an overall score of 0.7.

  FIG. 13 shows a system architecture 1300 for geotagging content in Sharepoint 2013®. Search index 1324 is one of a number of important components for enabling searching at Sharepoint 1302. Another important part that enables searching in Sharepoint 2013® 1302 is content capture for indexing content.

  The crawler 1304 crawls through different content sources 1306 and adds a list of metadata properties to each content. Content sources include, but are not limited to, for example, Sharepoint content, network file shares, or user or intranet content. The crawler 1304 is configured to securely connect to the content source 1306 and perform the function of associating documents from the source with their metadata as crawled properties. The crawler 1304 is configured to give the content a full crawl or incremental crawl. Crawled properties include, for example, but are not limited to, author, title, creation date, among others.

  Sharepoint 2013 (registered trademark) includes a content processing component 1308. The content processing component 1308 retrieves content from the crawler 1304 and prepares to index it. Content processing 1308 includes word processing (language detection), parsing, entity extraction management, content-based file format detection, content processing error reporting, natural language processing, and crawled properties to managed properties, among others. Includes mapping stage.

  Content processing 1308 is extended by the content enrichment web service (CEWS 1310). CEWS 1310 enables enrichment of content processing 1308 by allowing web service callout 1312 to call external web services to perform additional actions and enrich crawled data properties. Web service callout 1312 is a standard simple object access protocol (SOAP) request or other web service used to exchange crawled data structured information with entity enrichment service 1314. Call method. Web service callout 1312 includes a trigger condition configured to control when an external web service is called for enrichment processing in the content enrichment configuration object. The entity enrichment service 1314 also determines the document type of the crawled data and determines the content that comes in the form of images (scanned documents, pictures, etc.). When content in the form of an image is found, the entity enrichment service 1314 determines the location of the crawled document, such as, but not limited to, an OCR processing engine such as an optical character recognition component or other image processing component. Send to 1316. The OCR processing engine 1316 then retrieves and processes the image file and converts it asynchronously to a text file. The OCR processed file 1318 is then re-supplied to the crawler 1304, crawled as a text file, and sent back to the content processing 1308 for processing in the rest of the workflow.

  The system architecture 1300 includes an external geotagger web service 1320 and a named entity tagger service 1322. Geotagger web service 1320 and named entity tagger service 1322 are both software modules configured to function as web service application providers and to respond to web service callouts 1312. The geotagger web service 1320 identifies and disambiguates geographic entities from crawled content using natural language processing entity extraction techniques, machine learning models and other techniques. For example, the geotagger web service 1320 disambiguates geographic entities by analyzing the statistical co-occurrence of entities found in the gazetta. The geotagger web service 1320 includes a database of statistical co-occurrence entities that are linked to content found by the crawler 1304. Following that same technique, the named entity tagger service 1322 is used to extract additional entities or text features such as organizations, people or topics.

  The geo tagger web service 1320 analyzes the managed properties sent as input properties by the CEWS 1310 and identifies the geographic entities referenced in the text. Non-limiting examples of input properties include FileType, IsDocument, OriginalPath, and body, among others. The geotagger web service 1320 then geotags the text by creating or modifying managed properties with reference to each found geographic entity. The geotagger web service 1320 sends the modified or new managed property to the entity enrichment service 1314 where a transformation is performed to map the modified managed property and pass it to the CEWS 1310 as an output property. Return it. This same process is used to interact with the named entity tagger service 1322 for the extraction and entity tagging of other entities or other features such as organizations, people or topics.

  After the augmented managed property is returned by the entity enrichment service 1314, the property is merged with the crawled file management property and sent to the search index 1324.

  Once geographic and other entity tags are associated and indexed with content, search queries are performed using geographic or named entity features. Search UI 1326 in Sharepoint 2013® includes a specific display that assists the user in performing geographic based searches and supports improved display of faceted search results. The search UI 1326 can be a custom web part or can be done by modifying the standard layout of Sharepoint 2013® search with standard tools such as HTML, HTML5, JavaScript® and CSS. Also good.

  FIG. 14 is a flowchart 1400 illustrating the process steps for tagging content for a Sharepoint 2013® search. This process begins when the Sharepoint 2013® crawler component crawls content (step 1402). In one embodiment, the crawl is a full crawl and in another embodiment, the crawl is an incremental crawl. The crawler component then provides crawled properties and metadata to the content process (step 1404). A decision is made to verify whether the crawled content includes a geographic or named entity. For example, but not limited to, a trigger condition is used. The trigger condition includes a set of programming logic or rules that determine whether the content will benefit from geotagging or entity tagging. If the trigger condition evaluates to false, the crawled component is associated with the managed property (step 1406) and passed to the search index component (step 1408). If the trigger condition evaluates to true, the CEWS sends a web service callout to the entity enrichment service (step 1410). The entity enrichment service analyzes the sent content to determine if the content is in image format (scanned document, picture, etc.). The content found in the image format is processed asynchronously by the OCR engine and returned to be crawled as a text file by the crawling component (step 1412). If the content is not in an image format, the content is processed by a geotagging web server or name entity tagger service (step 1414). Web services extract and disambiguate geographic or named entities referenced in the content and enrich them with entity metadata. The identified entities and their metadata are returned as managed properties to the content processing component and associated with the content (step 1416). The associated metadata is then sent to the search index component (step 1406).

  While various aspects and embodiments have been disclosed, other aspects and embodiments are also contemplated. The various aspects and embodiments disclosed herein are for purposes of illustration and not limitation, and the true scope and spirit is indicated by the following claims.

  The foregoing method descriptions and process flow diagrams are presented by way of example only and are not intended to imply or imply that the steps of the various embodiments must be performed in the order presented. As will be apparent to those skilled in the art, the steps in the embodiments may be performed in any order. The words “then”, “next”, etc. do not limit the order of the steps, these words are simply used to guide the reader through the description of the method Only. Although the process flow diagram shows the operations as a series of processes, multiple operations can be performed in parallel or simultaneously. In addition, the order of operations may be reconfigured. Processes correspond to methods, functions, procedures, subroutines, subprograms, etc. When the process corresponds to a function, its termination corresponds to the return of the function to the calling function or the main function.

  The various exemplary logic blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be embodied as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described generally in terms of their functionality. Whether such a function is implemented as hardware or software depends on a specific application imposed on the entire system and design constraints. Those skilled in the art can implement the functions described herein in various ways for each specific application, but such implementation decisions should not be construed as departing from the scope of the present invention.

  Embodiments embodied in computer software are embodied in software, firmware, middleware, microcode, hardware description language, or a combination thereof. A code segment or machine-executable instruction represents a procedure, function, subprogram, program, routine, subroutine, module, software package, class, or instruction, data structure, or combination of program statements. A code segment is coupled to another code segment or a hardware circuit by passing and / or receiving information, data, arguments, parameters or memory contents. Information, arguments, parameters, data, etc. are passed, forwarded or transmitted via any suitable means including memory sharing, message passing, token passing, network transmission, etc.

  The actual software code or specialized control hardware used to implement these systems and methods is not intended to limit the invention. Accordingly, the operation and behavior of the system and method have been described without reference to specific software code, with the understanding that software and control hardware can be designed to implement the system and method based on the description herein. .

  When implemented in software, the functions are stored as one or more instructions or code on a non-transitory computer-readable or processor-readable storage medium. The steps of the method or algorithm disclosed herein are implemented in a processor-executable software module residing on a computer-readable or processor-readable storage medium. Non-transitory computer readable or processor readable media include both computer storage media and tangible storage media that facilitate transfer of a computer program from one place to another. A non-transitory processor readable storage medium is any available medium that can be accessed by a computer. By way of example, but not limitation, such non-transitory processor readable media may be RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage device, or It includes other tangible storage media used to store desired program code in the form of instructions or data structures and accessed by a computer or processor. As used herein, the term “disk & disc” includes a compact disc (CD), a laser disc (registered trademark), an optical disc, a digital diversity disc (DVD), a floppy disc, and a Blu-ray disc. The (disk) usually reproduces data magnetically, while the disc (disc) reproduces data optically with a laser. Combinations of the above are also included within the scope of computer-readable media. In addition, the operations of the method or algorithm may reside as one or a combination or set of codes and / or instructions on a non-transitory processor-readable medium and / or computer-readable medium that is incorporated into a computer program product.

  It will be apparent that the various components of the technology can be located in a remote part of the distributed network and / or the Internet, or in a dedicated secure, unsecure and / or encryption system. Thus, it will be apparent that the components of the system can be coupled to one or more devices or co-located at a particular node in a distributed network such as a telecommunications network. As is apparent from the above description, for computational efficiency reasons, the components of the system can be located anywhere in the distributed network without affecting the operation of the system. In addition, these components can be embedded in a dedicated machine.

  Further, the various links connecting the elements can be wired or wireless links or combinations thereof, or other known or later developed elements that can supply and / or communicate data to and from the connected elements. It is clear that there is. As used herein, the term module refers to any known or later developed hardware, software, firmware, or combination thereof that can perform the functions associated with the element. Also, the terms decision, computation and computing, and variations thereof, as used herein, are used interchangeably and encompass any type of method, process, mathematical operation or technique.

  The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Accordingly, the invention is not intended to be limited to the embodiments shown but is to be accorded the widest scope consistent with the claims and the principles and novel features disclosed herein.

  The embodiments described above are merely examples. Those skilled in the art will recognize numerous alternative components and embodiments that may be substituted for the specific examples described herein and still fall within the scope of the present invention.

DESCRIPTION OF SYMBOLS 100: Search system 102: Graphic user interface 104: Search engine 106: Server apparatus 108: Network connection 110: Entity extraction module 112: Entity co-occurrence knowledge base 114: Entity index type corpus 400: Search computer system 402: User interface 404: Search engine 406: Server device 408: Network connection 410: Entity extraction module 412: Fuzzy score matching module 414: Entity co-occurrence knowledge base database 700: Search system 702: User interface 704: Search engine 706: Server device 708: Network connection 710: Entity extraction module 712: Ambiguous Abnormal score matching module 714: Database of entity co-occurrence knowledge base 1000: Search system 1002: Search engine 1004: Entity database 1006: Trend database 1008: Search suggestion 1300: System architecture 1302: Sharepoint
1304: Crawler 1306: Content source 1308: Content processing 1312: Web service callout 1314: Entity enrichment service 1316: OCR processing engine 1318: OCR file 1320: Geotagger web service 1322: Named entity tagger service 1324: Search indexer 1326 : Search UI

Claims (56)

An entity extraction computer receives a search query including one or more entities from a client computer;
The entity extraction computer compares each entity with one or more co-occurrence of each entity in the co-occurrence database;
The entity extraction computer allows a subset of one or more entities from a search query to be shared by each entity in the subset based on the accuracy of the co-occurrence of that entity with one or more related entities in the electronic data corpus according to a co-occurrence database. In response to a determination that the reliability score of the origin database is exceeded,
An entity extraction computer assigns an index identifier (index ID) to each of the entities in the plurality of extracted entities;
An entity extraction computer saves an index ID for each of the plurality of extracted entities in an electronic data corpus, the electronic data corpus being indexed by an index ID corresponding to each of the one or more related entities. Yes,
Searching an entity indexed electronic data corpus to locate the plurality of extracted entities and to identify an index ID of a data record in which at least two of the plurality of extracted entities co-occur by a search server computer; And a search server computer builds a search result list having data records corresponding to the identified index ID;
A computer-implemented method comprising:   The method of claim 1, further comprising: classifying the search result list by relevance based on the reliability score by the search server computer; and transferring the classified search result list to the user device by the search server computer. Method.   The method of claim 1, wherein the plurality of extracted entities are ranked based on a confidence score.   The method of claim 1, wherein the entity extraction computer associates the extracted entities with one or more co-occurrence entities in an entity indexed electronic data corpus.   The method of claim 4, wherein the associated entities are ranked by a confidence score.   The method of claim 1, wherein each of the plurality of entities is selected from the group consisting of an individual, an organization, a geographical location, a date, and a time. Comprising one or more server computers having one or more processors executing computer readable instructions for a plurality of computer modules, comprising:
An entity extraction module configured to receive user input of search query parameters, the entity extraction module further comprising:
By comparing each entity in the plurality of extracted entities to an entity co-occurrence database that includes a confidence score representing the accuracy of the co-occurrence of the extracted entity and one or more related entities in the electronic data corpus, Extract multiple entities from search question parameters,
Specify an index identifier (index ID) for each entity in the plurality of extracted entities;
Saving an index ID for each of the plurality of extracted entities in an electronic data corpus, the electronic data corpus being indexed by an index ID corresponding to each of the one or more related entities;
And searching the entity indexed electronic data corpus to locate the plurality of extracted entities and identify an index ID of a data record in which at least two of the plurality of extracted entities co-occur A system further comprising a search server module configured to, wherein the search server module is further configured to build a search result list having data records corresponding to the identified index ID.   The system of claim 7, wherein the search server module is further configured to classify the search result list by relevance based on a reliability score and forward the sorted search result list to a user device.   The system of claim 7, wherein the plurality of extracted entities are ranked based on a confidence score.   The system of claim 7, wherein the entity extraction module is configured to associate the extracted entities with one or more co-occurrence entities in an entity indexed electronic data corpus.   The system of claim 10, wherein the associated entities are ranked by a confidence score.   The system of claim 7, wherein each of the plurality of entities is selected from the group consisting of an individual, an organization, a geographical location, a date, and a time. The entity extraction computer receives user input for search question parameters,
An entity co-occurrence database including a confidence score representing the accuracy of co-occurrence of the extracted entities and one or more related entities in the electronic data corpus with an entity extraction computer By comparing, extract multiple entities from search question parameters,
An entity extraction computer designates an index identifier (index ID) for each entity in a plurality of extracted entities;
An entity extraction computer saves an index ID for each of a plurality of extracted entities in an electronic data corpus, the electronic data corpus being indexed by an index ID corresponding to each of the one or more related entities. ,
A search server computer searches the entity indexed electronic data corpus to locate the plurality of extracted entities and to identify an index ID of a data record in which at least two of the plurality of extracted entities co-occur Building a search result list having data records corresponding to the identified index ID by the search server computer;
A non-transitory computer-readable medium having stored thereon computer-executable instructions.   The instructions further comprise classifying the search result list by relevance based on the reliability score by the search server computer and transferring the classified search result list to the user device by the search server computer. Item 14. A computer-readable medium according to Item 13.   The computer-readable medium of claim 13, wherein the plurality of extracted entities are ranked based on a confidence score.   The computer-readable medium of claim 13, wherein the instructions further associate an extracted entity with one or more co-occurring entities in an entity indexed electronic data corpus by an entity extraction computer.   The computer-readable medium of claim 16, wherein the associated entities are ranked by a confidence score.   The computer-readable medium of claim 13, wherein each of the plurality of entities is selected from the group consisting of an individual, an organization, a geographical location, a date, and a time. The entity extraction computer receives user input for search question parameters from the user interface,
The entity extraction computer compares the search query parameters with an entity co-occurrence database having co-occurrence instances of one or more entities in the electronic data corpus, and at least one corresponding to the one or more entities in the search query parameters Extract one or more entities from the search question parameters by identifying the entity type,
An ambiguity score matching computer selects an ambiguity matching algorithm that searches the entity co-occurrence database to identify one or more records associated with the search query parameter, the ambiguity matching algorithm comprising at least one identified entity type Corresponding to
An ambiguity score matching computer that searches the entity co-occurrence database using the selected ambiguity matching algorithm and forms one or more suggested search query parameters from one or more records based on the search; And the fuzzy score matching computer presents one or more suggested search query parameters via the user interface;
A method involving that.   21. The method of claim 19, further comprising searching the entity co-occurrence database by the fuzzy score matching computer using the selected fuzzy matching algorithm before user input is complete.   The method of claim 19, wherein the one or more records associated with the search query parameter include a conceptual feature.   The one or more suggested search question parameters include a plurality of suggested search question parameters, and the method further includes the fuzzy score matching computer to search the plurality of suggested search question parameters in user input. 20. The method of claim 19, further comprising classifying in descending order based on accessibility of matches to the query parameters.   23. The method of claim 22, wherein the fuzzy score matching computer presents the classified suggested search query parameters in a drop-down list via a user interface.   The method of claim 19, wherein the entity co-occurrence database is indexed.   The method of claim 19, wherein the entity co-occurrence database includes an entity-to-entity index.   The method of claim 19, wherein the entity co-occurrence database includes an entity-to-topic index.   20. The method of claim 19, wherein the entity co-occurrence database includes an entity-to-fact index. Comprising one or more server computers having one or more processors executing computer readable instructions for a plurality of computer modules, comprising:
An entity extraction module configured to receive user input of search query parameters from a user interface, the entity extraction module further comprising:
The search query parameter is compared to an entity co-occurrence database having co-occurrence instances of one or more entities in the electronic data corpus and at least one entity type corresponding to the one or more entities is identified in the search query parameter To extract one or more entities from the search query parameters,
And further configured as
An ambiguity score matching module configured to select an ambiguity matching algorithm that searches the entity co-occurrence database to identify one or more records associated with the search query parameter;
And the ambiguity matching module corresponds to at least one identified entity type, the ambiguity score matching module further comprising:
Search the entity co-occurrence database using the selected fuzzy matching algorithm and form one or more suggested search query parameters from one or more records based on the search, and 1 through the user interface A system configured to present one or more suggested search query parameters.   30. The system of claim 28, wherein the fuzzy score matching module is further configured to search an entity co-occurrence database using the selected fuzzy matching algorithm before user input is terminated.   30. The system of claim 28, wherein the one or more records associated with the search query parameter include conceptual features.   The one or more suggested search question parameters include a plurality of suggested search question parameters, and the fuzzy score matching computer further converts the plurality of suggested search question parameters to search question parameters in a user input. 29. The system of claim 28, wherein the system is configured to sort in descending order based on proximity of the matches.   36. The system of claim 32, wherein the fuzzy score matching computer is configured to present the classified plurality of suggested search query parameters via a user interface in a drop-down list.   30. The system of claim 28, wherein the entity co-occurrence database is indexed.   30. The system of claim 28, wherein the entity co-occurrence database includes an entity-to-entity index.   30. The system of claim 28, wherein the entity co-occurrence database includes an entity-to-topic index.   30. The system of claim 28, wherein the entity co-occurrence database includes an entity-to-fact index. An entity extraction computer receives user input of a partial search question parameter from a user interface, the partial search question parameter having at least one incomplete search question parameter;
The entity extraction computer compares the partial search query parameter with an entity co-occurrence database having co-occurrence instances of one or more first entities in the electronic data corpus, and one or more first in the partial search query parameter. Extracting one or more first entities from the partial search query parameters by identifying at least one entity type corresponding to an entity;
An ambiguity score matching computer selects an ambiguity matching algorithm that searches an entity co-occurrence database to identify one or more records associated with the partial search query parameter, the ambiguity matching algorithm comprising at least one identified entity Corresponding to the format,
An ambiguity score matching computer searches the entity co-occurrence database using the selected ambiguity matching algorithm and determines one or more first suggested search query parameters from one or more records based on the search. Forming,
An ambiguous score matching computer presenting one or more first suggested search query parameters via a user interface;
An entity extraction computer receives a user selection of one or more first suggested search query parameters to form a completed search query parameter;
An entity extraction computer to extract one or more second entities from the completed search query parameters;
An entity extraction computer searches an entity co-occurrence database to identify one or more entities associated with the one or more second entities to form one or more second suggested search query parameters And presenting one or more second suggested search query parameters via the user interface by the entity extraction computer;
A method involving that.   38. The method of claim 37, further comprising searching the entity co-occurrence database by the fuzzy score matching computer using the selected fuzzy matching algorithm before user input is complete.   38. The method of claim 37, wherein the one or more records associated with the partial search query parameter include conceptual features.   The one or more first suggested search query parameters include a plurality of first suggested search query parameters, and the method is further configured by the fuzzy score matching computer to provide the plurality of first suggested search query parameters. 38. The method of claim 37, further comprising classifying the searched query parameters in descending order based on the proximity of matches to the partial search query parameters in the user input.   41. The method of claim 40, wherein the fuzzy score matching computer presents the categorized first suggested search query parameters in a drop-down list via a user interface.   40. The method of claim 39, wherein the entity co-occurrence database is indexed.   38. The method of claim 37, wherein the entity co-occurrence database includes an entity-to-entity index.   38. The method of claim 37, wherein the entity co-occurrence database includes an entity-to-topic index.   38. The method of claim 37, wherein the entity co-occurrence database includes an entity-to-fact index. Comprising one or more server computers having one or more processors executing computer readable instructions for a plurality of computer modules, comprising:
An entity extraction module configured to receive user input of partial search question parameters from a user interface, the partial search question parameters having at least one unfinished search question parameter; Furthermore,
Comparing the partial search query parameter to an entity co-occurrence database having co-occurrence instances of one or more first entities in the electronic data corpus and corresponding to at least one first entity in the partial search query parameters Extracting one or more first entities from the partial search query parameters by identifying one entity type;
And further configured
An ambiguity score matching module configured to select an ambiguity matching algorithm that searches the entity co-occurrence database to identify one or more records associated with the partial search query parameter, the ambiguity matching algorithm comprising at least one Corresponding to two identified entity types, and its fuzzy score matching module further comprises:
Search the entity co-occurrence database using the selected fuzzy matching algorithm and form one or more first suggested search query parameters from one or more records based on the search; and Presenting one or more first suggested search query parameters via an interface;
The entity extraction module is further configured as
Receiving a user selection of one or more first suggested search query parameters to form a completed search query parameter;
Extracting one or more second entities from the completed search query parameters;
Search the entity co-occurrence database to identify one or more entities associated with the one or more second entities and form one or more second suggested search query parameters; and Via which one or more second suggested search query parameters are presented,
Configured system.   47. The system of claim 46, wherein the fuzzy score matching module is further configured to search an entity co-occurrence database using the selected fuzzy matching algorithm before user input ends.   48. The system of claim 46, wherein the one or more records associated with the partial search query parameter include conceptual features.   The one or more first suggested search question parameters include a plurality of first suggested search question parameters, and the fuzzy score matching computer further includes the plurality of first suggested search question parameters. 47. The system of claim 46, wherein the system is configured to classify parameters in descending order based on accessibility of matches to partial search query parameters in user input.   50. The system of claim 49, wherein the fuzzy score matching computer is configured to present the categorized first suggested search query parameters in a drop-down list via a user interface.   48. The system of claim 46, wherein the entity co-occurrence database is indexed.   48. The system of claim 46, wherein the entity co-occurrence database includes an entity-to-entity index.   47. The system of claim 46, wherein the entity co-occurrence database includes an entity-to-topic index.   48. The system of claim 46, wherein the entity co-occurrence database includes an entity-to-fact index. A computer receives a search query containing one or more data strings from a search engine, each entity corresponding to a subset of one or more strings,
Identifying one or more entities in one or more data strings based on comparing one or more entities against an entity database and a trend database by a computer;
Identifying one or more features in one or more data strings not identified by the computer as corresponding to at least one entity;
The computer assigns each of the one or more features to at least one of the one or more entities based on a matching algorithm;
The computer specifies an extraction score for each entity based on the score specified for each feature specified for each entity,
A computer receives a first search list from an entity database that includes one or more entities having a score that is within a threshold distance from an extraction score for each entity;
A computer receives a second search list from the trend database that includes one or more entities having a score that is within a threshold distance from the extracted score of each entity;
A computer generates a grand total list including a first search list and a second search list, wherein the entities of the total list are ranked according to the score of each total list, and are suggested by the computer according to the total list. Give a search,
A computer-implemented method comprising: The computer receives a plurality of data streams each associated with a plurality of data sources,
The computer generates an array of properties associated with each data stream,
In response to the computer detecting a trigger condition associated with the data in the data stream,
The computer generates geographic data related to the data stream data,
In response to the computer not detecting the data source trigger condition,
In response to mapping by the computer an array of properties for the data source to a set of managed properties associated with the search index and determining that the content type of the data source is image data,
An optical character recognition routine is performed on metadata associated with data received from the data source by the computer, and an updated data stream from the data source from the web service identified by the metadata by the computer And the data source is associated with the web service identified in the metadata,
Computer-implemented method.

Images