Method and apparatus for performing semantic-based data analysis

• US 9,298,818 B1
• Filed: 05/31/2011
• Issued: 03/29/2016
• Est. Priority Date: 05/28/2010
• Status: Active Grant

• Current Assignee: SRI International, Inc.
• Original Assignee: SRI International, Inc.
• Inventors: Donneau-Golencer, Thierry, Nitz, Kenneth, Dunkley, David
• Primary Examiner(s): Lee, Wilson
• Application Number:
  US13/149,319
• Time in Patent Office: 1,764 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 16/68 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually

Abstract

A computer implemented method and apparatus for analyzing content of a plurality of documents. The method extracts semantic information from content of a plurality of documents; assigns weights to the semantic information; assigns links between documents containing similar semantic information; assigns a weight to each link; extracts information about the content of the plurality of documents by using the weighted links and weighted semantics to cluster the documents, perform inferential analysis, or both.

Claims

1. A computer implemented method for performing a semantic-based analysis of a plurality of documents with regard to a query, the method comprising:

extracting, by a content processor, semantic information from content of a plurality of documents, the semantic information including a plurality of factual beliefs;

assigning a weight to each of the factual beliefs, the weights being based at least partly on relevance to the query;

assigning links between documents containing related semantic information;

assigning a weight to each link indicating a confidence level that the documents comprise semantically related information; and

determining one or more inferential relationships regarding the query based upon analysis of the plurality of weighted factual beliefs and the links, wherein determining one or more inferential relationships includes applying rules to the weighted factual beliefs and the weighted links to obtain the one or more inferential relationships.

2. The method of claim 1 further comprising accessing semantic characteristics to define the semantic information to be extracted, wherein the semantic characteristics comprise a semantic type and a semantic feature, and the semantic feature is a specific instance of the semantic type.

3. The method of claim 1, wherein the weight assigned to each of the factual beliefs is further based on a frequency of terms associated with the factual beliefs found in the documents.

4. The method of claim 1, wherein the inferring rules and the weights assigned to the links are based on least partly on:

proximity of semantics within text of a document, semantics belonging to a mutual collection, proximity of time and geographic location of semantics, and similar verb groupings for semantics.

5. The method of claim 3, further comprising determining at least one inferential relationship among the extracted semantic information from at least two different documents, wherein the at least one inferential relationship is not explicitly stated in the contents of the at least two different documents, and representing the at least one relationship in an inference graph.

6. The method of claim 1 wherein the weights of the assigned links are based at least in part on the semantic information.

7. A computer implemented method for performing a semantic-based analysis of a plurality of documents with regard to a query, the method comprising:

extracting, by a content processor, semantic information from content of a plurality of documents, the semantic information including a plurality of factual beliefs;

assigning a weight to each of the factual beliefs, the weights being based at least partly on relevance to the query;

assigning links between documents containing related semantic information; and

assigning a weight to each link indicating a confidence level that the documents comprise semantically related information;

performing clustering of the plurality of documents by applying a cluster tool to identify documents having content with common information;

adjusting at least one weight of at least one of the weighted factual beliefs or the weighted links to modify at least one cluster of documents; and

inferring one or more inferential relationships regarding the query based upon analysis of the plurality of weighted factual beliefs and the links.

8. The method of claim 7, wherein adjusting the at least one weight further comprises applying feedback to the factual belief weights and the link weights, to thereby improve a confidence level of the inferential relationships.

9. An apparatus for performing a semantic-based analysis of a plurality of documents with regard to a query, the apparatus comprising:

a central processing unit (CPU);

a content processor for extracting semantic information from content of a plurality of documents, the semantic information including a plurality of factual beliefs, assigning a weight to each of the factual beliefs, the weights being based at least partly on relevance to the query, and assigning links between documents containing related semantic information;

a cluster identifier that performs clustering of the plurality of documents by applying a cluster tool to identify documents having content with common information, wherein the cluster identifier adjusts at least one weight of at least one of the weighted factual beliefs or the weighted links to modify at least one cluster of documents; and

an inference identifier that determines one or more inferential relationships regarding the query based upon analysis of the plurality of weighted factual beliefs and the links.

10. The apparatus of claim 9 wherein the semantic information to be extracted is defined by semantic characteristics, wherein the semantic characteristics comprise a semantic type and a semantic feature, and the semantic feature is a specific instance of the semantic type.

11. The apparatus of claim 9, wherein the weight assigned to each of the factual beliefs is further based on a frequency of terms associated with the factual beliefs and wherein the inference identifier applies inferring rules to the weighted semantic information and weighted links.

12. The apparatus of claim 11, wherein the content processor further assigns a weight to each link indicating a confidence level that the documents comprise semantically related information, and

wherein the inferring rules and the weights assigned to the links are based at least partly on:

proximity of semantics within text of a document, semantics belonging to a mutual collection, proximity of time and geographic location of semantics, and similar verb groupings for semantics.

13. The apparatus of claim 9 wherein the inference identifier determines at least one inferential relationship among the extracted semantic information from at least two different documents, wherein the at least one inferential relationship is not explicitly stated in the contents of the at least two different documents, and representing the at least one relationship in an inference graph.

14. The apparatus of claim 11 wherein the content processor defines weights of the assigned links based on the semantic information.

15. The apparatus of claim 9 wherein adjusting the at least one weight further apply feedback to the factual belief weights and the link weights, to thereby improve a confidence level of the one or more inferential relationships.

16. A non-transitory computer readable medium for storing computer instructions that, when executed by a processor, cause the processor to perform a method for performing a semantic-based analysis of a plurality of documents with regard to a query, the method comprising:

extracting semantic information from content of a plurality of documents, the semantic information including a plurality of factual beliefs;

assigning a weight to each of the factual beliefs, the weights being based at least partly on relevance to the query;

assigning links between documents containing related semantic information;

assigning a weight to each link indicating a confidence level that the documents comprise semantically related information; and

determining one or more inferential relationships regarding the query based upon analysis of the plurality of weighted factual beliefs and the links, wherein determining one or more inferential relationships includes applying rules to the weighted factual beliefs and the weighted links to obtain the one or more inferential relationships.