Annotating entities using cross-document signals

• US 9,465,865 B2
• Filed: 08/16/2012
• Issued: 10/11/2016
• Est. Priority Date: 05/29/2012
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: De, Sushovan, Singh, Amit K., Visweswariah, Karthik
• Primary Examiner(s): Baderman, Scott
• Assistant Examiner(s): EDWARDS, JASON T
• Application Number:
  US13/587,011
• Publication Number:
  US 20130325849A1
• Time in Patent Office: 1,517 Days
• Field of Search: 715/230, 715/737, 707/737
• US Class Current: 1/1
• CPC Class Codes: G06F 16/355 Class or cluster creation or modification

Abstract

Techniques for annotating an entity in a document corpus using cross-document signals. A method includes determining which documents in a document corpus mention an entity of interest, clustering the documents that mention an entity of interest according to a temporal signal, a structural signal and/or a content signal, thereby forming at least one cluster of documents, and annotating at least one document in the at least one cluster of documents by marking each occurrence of the entity in the at least one document.

Claims

1. A method for annotating an entity in a document corpus using cross-document signals, the method comprising:

determining which documents in a document corpus of multiple documents mention an entity of interest;

clustering the documents that mention an entity of interest according to similarities across a temporal signal, a structural signal and a content signal, thereby forming multiple clusters of documents;

annotating each document in the multiple clusters of documents with an annotation by marking each occurrence of the entity in each document;

calculating a confidence measure for each occurrence of the entity in each document in each of the multiple clusters, wherein said confidence measure comprises the sum of (i) a measure of similarity between the given occurrence of the entity and the entity of interest, and (ii) a measure of similarity between the documents within the cluster of the given document via

2. The method of claim 1, comprising marking an occurrence of the entity in at least one document as not applicable, in order to opt out of at least one of an incorrect and uncertain annotation.

3. The method of claim 1, comprising applying a model to each cluster that provides weights to features of a signal to guide said annotating.

4. The method of claim 1, wherein a temporal signal corresponds to a situation in which a document is from the same time epoch as another document.

5. The method of claim 4, wherein the time epoch is measured via at least one granularity including at least one of minutes, hours, days, weeks, and months.

6. The method of claim 1, wherein a structural signal corresponds to a situation in which a document is part of a larger document arrangement.

7. The method of claim 1, wherein said determining comprises using a dictionary of entities.

8. The method of claim 7, wherein the dictionary contains a description corresponding to each entity.

9. The method of claim 1, comprising training a set of documents with labeled entities to determine at least one of a threshold for clustering documents and a document similarity weight.

10. The method of claim 1, comprising annotating a single entity in a single document.

11. The method of claim 10, comprising determining an entity-to-context match for a mention of the entity from the document in an entity database without considering a signal from other documents or other entities in the document.

12. The method of claim 1, comprising annotating multiple entities in a single document.

13. The method of claim 12, comprising determining an entity-to-context match for each mention of the multiple entities from the document in an entity database without considering a signal from other documents.

14. The method of claim 1, comprising annotating a single entity in multiple documents.

15. The method of claim 14, comprising determining at least one of an entity-to-context match and a document-to-document match for a mention of the entity from the document in an entity database, taking into account similarity and temporal proximity to other documents without considering a signal from other entities in the document.

16. The method of claim 1, comprising annotating multiple entities in multiple documents.

17. The method of claim 16, wherein said determining includes considering signals from multiple entities in each document as well as temporal and textual similarity to other documents in the corpus.

18. A method for annotating an entity in a document corpus, the method comprising:

processing each document in a corpus of multiple documents obtained from at least one online source to identify which documents mention an entity of interest;

using a description of the entity derived from a database to generate at least one context feature for the entity;

processing each of the documents that mention the entity of interest by comparing text from each document with the at least one context feature and grouping the documents with a comparison similarity above a pre-determined threshold into a cluster of documents;

annotating each document in the cluster of documents with an annotation by marking each occurrence of the entity in each document;

calculating a confidence measure for each occurrence of the entity in each document in the cluster of documents, wherein said confidence measure comprises the sum of (i) the comparison similarity between the given occurrence of the entity and the entity of interest, and (ii) a measure of similarity between the documents within the cluster of documents via