Knowledge-based entity detection and disambiguation

• US 9,665,643 B2
• Filed: 12/30/2011
• Issued: 05/30/2017
• Est. Priority Date: 12/30/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Zhou, Yiping, Lv, Zhengdong, Cao, Yohn, Li, Kang, Li, Yi
• Primary Examiner(s): NGUYEN, MERILYN P
• Application Number:
  US13/340,722
• Publication Number:
  US 20130173604A1
• Time in Patent Office: 1,978 Days
• Field of Search: 707723
• CPC Class Codes: G06F 16/3346 using probabilistic model
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9538 Presentation of query results

Abstract

An entity-based search system is described herein that detects and recognizes entities in Internet-based content and uses this recognition to organize search results. The system associates one or more entity identifiers with a web page and stores this information as metadata of the page in a search engine index. This metadata will enable entity-based queries as well as rich data presentations in a search engine result page (SERP), including grouping results by entities, filtering results by one or more particular entities, or re-ranking search results based on user preference of entities. Thus, the entity-based search system allows users to identify a particular entity the user is interested in finding, and to receive search results directly related to that entity.

Claims

1. A computer-implemented method to identify search results based on entity-based knowledge, the method comprising:

receiving a search query that includes one or more entities;

enumerating one or more entities from an entity-based knowledge store that may match entities in the received query;

identifying one or more potentially matching entities from the enumerated entities, wherein a matching entity is one that may be a target of the query;

ranking the identified potentially matching entities to distinguish a relative level of match between each identified entity and one or more entities in the received query;

selecting one or more of the identified potentially matching entities as matching entities with which to go forward with a search;

performing an entity-based search that narrows results by applying one or more top ranked, selected matching entities;

ranking multiple results returned from the entity-based search so as to order the results by relevance to the received search query, by (i) dynamically determining entity identifiers by;

identifying a first group of one or more results each without an entity identifier;

identifying a second group of one or more results, different from the first group, each with an entity identifier; and

comparing members of the first group with members of the second group, and then (ii) grouping results by entities with which the results are associated; and

providing the ranked search results in response to the received query,

wherein the preceding steps are performed by at least one processor.

2. The method of claim 1 wherein receiving the search query comprises receiving the query directly through a user interface.

3. The method of claim 1 wherein receiving the search query comprises receiving additional information related to a user'"'"'s purpose in submitting the query.

4. The method of claim 1 wherein enumerating entities comprises accessing potentially matching entities based on query keywords or natural language analysis to score a relationship between the query and known entities in the knowledge store.

5. The method of claim 1 wherein identifying potentially matching entities comprises performing one or more initial steps to reduce the number of entities from the entire set of the knowledge store down to a subset relevant to the received query.

6. The method of claim 1 wherein ranking potentially matching entities comprises separating multiple potential entities in the query and identifying a ranked set of matches for each potential entity.

7. The method of claim 1 wherein ranking potentially matching entities comprises considering the query as a whole and rank a set of matches for all of the entities in the query.

8. The method of claim 1 wherein selecting matching entities for search comprises displaying a user interface to a user for disambiguating which among a candidate set of entities the user intended for the search.

9. The method of claim 1 wherein selecting matching entities for search comprises determining that the level of match for a highest ranked entity exceeds a threshold sufficient to automatically identify that entity as the matching entity for search.

10. The method of claim 1 wherein performing the entity-based search comprises applying the identified entities as well as any additional information in the query to find search results relevant for the identified entities.

11. The method of claim 1 wherein performing the entity-based search comprises leveraging a reverse index created during discovery of entity information, which indexes from a known entity to a set of content results that are most relevant for that known entity.

12. The method of claim 1 wherein performing the entity-based search comprises further refining the search results by finding the pages among those that also match any additional words or information in the currently received query.

13. The method of claim 1, wherein, in the providing, the ranked search results are provided in respective lists that reflect the grouping so that a respective result group is followed by an associated result list.

14. The method of claim 1, wherein the performing an entity-based search that narrows results by applying one or more top ranked, selected matching entities includes asking a user for a threshold level of matching to apply in the ranking.

15. The method of claim 1, wherein the providing the ranked search results in response to the received query includes providing a disambiguation control with a user interface that receives, from a user, explicit identification of an entity for which information is sought.

16. A computer system for knowledge-based entity detection and disambiguation, the system comprising

a processor and memory configured to execute software instructions embodied within the following components:

a knowledge repository component that stores information related to previously discovered entities and relationships between entities;

a repository update component that adds and updates entity information stored in the knowledge repository component;

an entity detection component that receives discovered content and detects one or more entities embedded within the discovered content;

an entity-based indexing component that creates and maintains an index between detected entities and content related to the detected entities;

an entity enumeration component that receives a user query and enumerates previously detected entities that potentially match the user query;

an entity disambiguation component

classifies and ranks enumerated entities that potentially match the user query to qualify a relative level of match between various entities to the user query;

automatically select one or more top-ranked enumerated entities that match the user query based on information in the user query;

an entity-based retrieval component that provides an interface through which a user can submit queries to search for information based on entities and receive entity-based search results; and

a result-ranking component that ranks the entity-based search results provided to the user, by (i) dynamically determining entity identifiers by;

identifying a first group of one or more results each without an entity identifier;

identifying a second group of one or more results, different from the first group, each with an entity identifier; and

comparing members of the first group with members of second group, and then (ii) grouping results by entities with which the results are associated.

17. The system of claim 16 wherein the repository update component is invoked to add new entities and relationships that are discovered during crawling, and to update changed or removed entities.

18. The system of claim 16 wherein the entity detection component invokes the repository update component for each discovered entity to store discovered entity information for later matching to user queries.

19. The system of claim 16 wherein the entity detection component operates on a variety of types of content, including textual and audiovisual data.

20. The system of claim 16 wherein the entity-based index component provides a reverse index that maps entity identifiers to content pages related to a particular entity identifier.

21. The system of claim 16 wherein the entity disambiguation component outputs a list of matching entities ordered by level of match from which the user can select an entity representative of the user'"'"'s present search target.

22. The system of claim 16, wherein the entity-based indexing component creates and maintains the index between detected entities and content related to the detected entities by (i) associating entity identifiers that each unit of input text contains with the content and (ii) storing information of the associating as metadata of the content, in the index.

23. A computer-readable storage medium comprising instructions for controlling a computer system to discover entities in a corpus of content, wherein the instructions, upon execution, cause a processor to perform actions comprising:

receiving a search query that includes one or more entities;

enumerating one or more entities from an entity-based knowledge store that may match entities in the received query;

identifying one or more potentially matching entities from the enumerated entities, wherein a matching entity is one that may be a target of the query;

ranking the identified potentially matching entities to distinguish a relative level of match between each identified entity and one or more entities in the received query;

selecting one or more of the identified potentially matching entities as matching entities with which to go forward with a search;

performing an entity-based search that narrows results by applying one or more of the top ranked, selected matching entities;

ranking multiple results returned from the entity-based search so as to order the results by relevance to the received search query, by (i) dynamically determining entity identifiers by;

identifying a first group of one or more results each without an entity identifier;

identifying a second group of one or more results, different from the first group, each with an entity identifier; and

comparing members of the first group with members of the second group, and then (ii) grouping results by entities with which the results are associated; and

providing the ranked search results in response to the received query.