TAGGING ENTITIES WITH DESCRIPTIVE PHRASES

• US 20130132381A1
• Filed: 11/17/2011
• Published: 05/23/2013
• Est. Priority Date: 11/17/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Chakrabarti, Kaushik, Chaudhuri, Surajit, Cheng, Tao
• Granted Patent: US 9,298,825 B2
• Time in Patent Office: Days
• US Class Current: 707/730
• CPC Class Codes: G06F 16/532 Query formulation, e.g. graphical querying
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

A plurality of description phrases associated with a first domain may be determined, based on an analysis of a first plurality of documents to determine co-occurrences of the description phrases with one or more name labels associated with the first domain. An entity associated with the first domain may be obtained. An analysis of a second plurality of documents may be initiated to identify co-occurrences of mentions of the obtained entity and one or more of the plurality of description phrases, and contexts associated with each of the co-occurrences of the mentions and description phrases, in each one of the second plurality of documents. A description tag association between the obtained entity and one of the description phrases may be determined, based on an analysis of the identified contexts.

Claims

1. A system comprising:

an etag collector configured to obtain a plurality of description indicators associated with a first domain;

an entity receiving component configured to obtain an entity indicator associated with the first domain;

a context determination component configured to initiate, via a device processor, an analysis of a plurality of documents to identify occurrences of mentions of the obtained entity indicator and contexts associated with each of the occurrences of the mentions in each one of the plurality of documents;

a proximity determination component configured to determine proximities of the description indicators to the occurrences of mentions of the obtained entity indicator, for each one of the plurality of documents that includes one or more of the description indicators and one or more occurrences of mentions of the obtained entity indicator, based on the obtained contexts; and

an etag association component configured to generate a description tag association between the obtained entity indicator and one of the description indicators, based on the determined proximities.

2. The system of claim 1, wherein:

the etag collector is configured to obtain the plurality of description indicators associated with the first domain based on determining the description indicators based on lexical patterns that include one of the description indicators and a name associated with the first domain.

3. The system of claim 1, wherein:

the description indicators include one or more attributes associated with one or more entities that represent instances of items included in the first domain.

4. The system of claim 1, wherein:

the context determination component is configured to initiate the analysis of the plurality of documents to identify occurrences of mentions of the obtained entity indicator and contexts associated with each of the occurrences of the mentions in each one of the plurality of documents, based on identifying co-occurrences of the mentions of the obtained entity indicator within a proximity window in text included in each one of the plurality of documents.

5. The system of claim 1, further comprising:

a fitness evaluator configured to determine a measure of fitness associated with the description tag association based on a goodness of fit value, determined in accordance with;

6. The system of claim 1, wherein:

the proximity determination component is configured to determine proximities of the description indicators to the occurrences of mentions of the obtained entity indicator, based on a number of tokens between each of the description indicators and occurrences of mentions of the obtained entity indicator.

7. The system of claim 1, further comprising:

an entity-specific determination component configured to determine whether one or more of the plurality of documents includes a significant percentage of text that is focused on one or more mentions of the obtained entity indicator.

8. The system of claim 1, further comprising:

a token determination component configured to determine one or more tokens associated with the obtained entity indicator, the one or more tokens included in one or more name variants associated with the obtained entity indicator.

9. The system of claim 8, further comprising:

a token weight assignment component configured to assign a weight to each respective one of the tokens associated with the obtained entity indicator.

10. The system of claim 9, wherein:

the context determination component is configured to initiate the analysis of the plurality of documents to identify occurrences of mentions of the obtained entity indicator and contexts associated with each of the occurrences of the mentions in each one of the plurality of documents, the identification of the mentions based on an analysis of Jaccard containment of subsequences of tokens included in a set of web documents, via a MapReduce platform.

11. The system of claim 8, further comprising:

an entity index generation component configured to generate an inverted index based on the one or more name variants associated with the obtained entity indicator,

wherein identifying occurrences of mentions of the obtained entity indicator is based on accessing the inverted index.

12. The system of claim 11, wherein:

the entity index generation component is configured to generate the inverted index based on the one or more name variants associated with the obtained entity indicator, based on prefix filtering.

13. The system of claim 11, wherein:

the entity index generation component is configured to generate the inverted index based on the one or more name variants associated with the obtained entity indicator, based on;

individual token inverted index entries, or

tokenset inverted index entries.

14. A method comprising:

determining a plurality of description phrases associated with a first domain, based on an analysis of a first plurality of documents;

obtaining an entity associated with the first domain;

initiating an analysis of a second plurality of documents to identify occurrences of mentions of the obtained entity and contexts associated with each of the occurrences of the mentions in each one of the second plurality of documents; and

generating, via a device processor, a description tag association between the obtained entity and one of the description phrases, based on the identified occurrences of mentions of the obtained entity and contexts associated with each of the occurrences of the mentions.

15. The method of claim 14, further comprising:

determining proximities of the description phrases to the occurrences of mentions of the obtained entity, for each one of the second plurality of documents that includes one or more of the description phrases and one or more of the occurrences of mentions of the obtained entity, based on the obtained contexts, wherein

generating the description tag association includes generating the description tag association between the obtained entity and one of the description phrases, based on the determined proximities.

16. The method of claim 15, wherein:

determining proximities of the description phrases to the occurrences of mentions of the obtained entity includes;

determining a number of tokens separating the respective description phrases from the occurrences of mentions of the obtained entity included in subsequences of tokens included in each one of the second plurality of documents that includes one or more of the description phrases and one or more of the occurrences of mentions of the obtained entity.

17. A computer program product tangibly embodied on a computer-readable medium and including executable code that causes at least one data processing apparatus to:

determine a plurality of description phrases associated with a first domain, based on an analysis of a first plurality of documents to determine co-occurrences of the description phrases with one or more name labels associated with the first domain;

obtain an entity associated with the first domain;

initiate an analysis of a second plurality of documents to identify co-occurrences f mentions of the obtained entity and one or more of the plurality of description phrases, and contexts associated with each of the co-occurrences of the mentions and description phrases, in each one of the second plurality of documents; and

determine a description tag association between the obtained entity and one of the description phrases, based on an analysis of the identified contexts.

18. The computer program product of claim 17, wherein:

the first plurality of documents and the second plurality of documents are each included in a set of web documents, wherein

initiating the analysis of the second plurality of documents to identify co-occurrences f mentions of the obtained entity and one or more of the plurality of description phrases, and contexts associated with each of the co-occurrences of the mentions and description phrases, in each one of the second plurality of documents includes initiating the analysis via a MapReduce platform.

19. The computer program product of claim 17, wherein the executable code is configured to cause the at least one data processing apparatus to:

determine proximities of the description phrases to the occurrences of mentions of the obtained entity, for each one of the second plurality of documents that includes one or more of the description phrases and one or more of the occurrences of mentions of the obtained entity, based on the obtained contexts, wherein

generating the description tag association includes generating the description tag association between the obtained entity and one of the description phrases, based on the determined proximities.

20. The computer program product of claim 17, wherein the executable code is configured to cause the at least one data processing apparatus to:

determine the plurality of description phrases associated with the first domain, based on an analysis of a first plurality of documents to determine co-occurrences of the description phrases with one or more name labels associated with the first domain, based on a plurality of text variants of Hearst patterns; and

initiate, via a MapReduce platform, the analysis of the second plurality of documents to identify co-occurrences of the mentions of the obtained entity and one or more of the plurality of description phrases, and contexts associated with each of the co-occurrences of the mentions and description phrases, an identification of the mentions based on an analysis of Jaccard containment of subsequences of tokens included in a set of web documents.