Question answering system, data search method, and computer program

• US 20060277165A1
• Filed: 12/14/2005
• Published: 12/07/2006
• Est. Priority Date: 06/03/2005
• Status: Active Grant

• Current Assignee: Fujifilm Business Innovation Corp. (Fujifilm Holdings Corporation)
• Original Assignee: Fuji Xerox Company Limited (Fujifilm Holdings Corporation)
• Inventors: Masuichi, Hiroshi, Yoshimura, Hiroki, Sugihara, Daigo, Ohkuma, Tomoko
• Granted Patent: US 7,418,443 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation

Abstract

A question answering system includes a question answering unit, an analysis unit, a tree structure generation unit, a feature extraction unit, an evaluation unit and a re-ranking unit. The question answering unit executes search processing based on an input question and acquires answer candidates to the question. The analysis unit executes syntactic analysis processing or syntactic and semantic analysis processing on a passage obtained as a result of the search processing. The tree structure generation unit generates a tree structure based on an analysis result. The feature extraction unit extracts a relation between a search word applied in the search processing and each acquired answer candidate as a feature corresponding to each answer candidate from the tree structure. The evaluation unit determines an evaluation value of each answer candidate based on the feature extracted. The re-ranking unit re-ranks the answer candidates based on the evaluation values.

Claims

1. A question answering system comprising:

a question answering unit that executes search processing on a basis of an input question and acquires answer candidates to the question;

an analysis unit that executes syntactic analysis processing or syntactic and semantic analysis processing on a passage obtained as a result of the search processing executed by the question answering unit;

a tree structure generation unit that generates a tree structure on a basis of an analysis result of the analysis unit;

a feature extraction unit that extracts a relation between a search word applied in the search processing executed by the question answering unit and each acquired answer candidate as a feature corresponding to each answer candidate from the tree structure generated by the tree structure generation unit;

an evaluation unit that determines an evaluation value of each answer candidate on a basis of the feature extracted by the feature extraction unit; and

a re-ranking unit that re-ranks the answer candidates on a basis of the evaluation values determined by the evaluation unit.

2. The question answering system according to claim 1, wherein the evaluation unit executes evaluation processing, to which a machine learning technique is applied, on the feature extracted by the feature extraction unit to determine the evaluation value of each answer candidate.

3. The question answering system according to claim 1, wherein:

the analysis unit executes the syntactic and semantic analysis processing on the passage to generate an f-structure holding a case relation between words of the passage to be analyzed,

the tree structure generation unit generates a tree structure based on predicate-argument structure explicitly expressing the case relation on a basis of the f-structure, and

the feature extraction unit executes processing of extracting the case relation between the search word and each answer candidate as the feature corresponding to each answer candidate on a basis of the tree structure.

4. The question answering system according to claim 1 wherein:

the analysis unit executes the syntactic analysis processing on the passage,

the tree structure generation unit generates a syntactic tree on a basis of a result of the syntactic analysis processing executed by the analysis unit, and

the feature extraction unit executes processing of extracting number of predicates between the search word and each answer candidate as the feature on a basis of the syntactic tree.

5. The question answering system according to claim 1, wherein the evaluation unit comprises:

a plurality of evaluation unit that execute evaluation processing, to which different machine learning techniques are applied, on the feature extracted by the feature extraction unit; and

an evaluation integration unit that integrates a plurality of evaluation results of the plurality of evaluation unit to determine a final evaluation value of each answer candidate.

6. The question answering system according to claim 1 further comprising:

a tree structure database that stores data of the tree structure generated by the tree structure generation unit in association with the passage based on which the data of the tree structure is generated; and

a passage search unit that executes search processing of the tree structure database on a basis of the passage input from the question answering unit, wherein;

if the search processing executed by the passage search unit finds a passage in the tree structure database, the tree structure generation processing executed by the tree structure generation unit is skipped and the tree structure stored in association with the passage found from the tree structure database is used.

7. A data search method comprising:

executing search processing on a basis of an input question and acquires answer candidates to the question;

executing syntactic analysis processing or syntactic and semantic analysis processing on a passage obtained as a result of the search processing;

generating a tree structure on a basis of a result of the analysis;

extracting a relation between a search word applied in the search processing and each acquired answer candidate as a feature corresponding to each answer candidate from the tree structure generated;

determining an evaluation value of each answer candidate on a basis of the feature extracted; and

re-ranking the answer candidates on a basis of the evaluation values determined.

8. The data search method according to claim 7, wherein the determining of the evaluation value comprises executing evaluation processing, to which a machine learning technique is applied, on the feature extracted to determine the evaluation value of each answer candidate.

9. The data search method according to claim 7, wherein:

the executing of the syntactic analysis processing or the syntactic and semantic analysis processing comprises executing the syntactic and semantic analysis processing on the passage to generate an f-structure holding a case relation between words of the passage to be analyzed,

the generating of the tree structure comprises generating a tree structure based on predicate-argument structure explicitly expressing the case relation on a basis of the f-structure, and

the extracting comprises executing processing of extracting the case relation between the search word and each answer candidate as the feature corresponding to each answer candidate on a basis of the tree structure.

10. The data search method according to claim 7, wherein:

the executing of the syntactic analysis processing or the syntactic and semantic analysis processing comprises executing the syntactic analysis processing on the passage,

the generating of the tree structure comprises generating a syntactic tree on a basis of a result of the syntactic analysis processing, and

the extracting comprises executing processing of extracting number of predicates between the search word and each answer candidate as the feature on a basis of the syntactic tree.

11. The data search method according to claim 7, wherein the determining of the evaluation value comprises:

executing a plurality of evaluation processings, to which different machine learning techniques are applied, on the feature extracted; and

integrating a plurality of evaluation results to determine a final evaluation value of each answer candidate.

12. The data search method according to claim 7 further comprising:

searching a tree structure database storing data of the tree structure in association with the passage based on which the data of the tree structure is generated with using the passage acquired as queries; and

if the searching of the tree structure database finds a passage in the tree structure database, executing feature extraction processing using a tree structure stored in association with the passage found from the tree structure database.

13. A computer program stored in a recording medium, the computer program causing a computer to execute a data search processing comprising:

executing search processing on a basis of an input question and acquires answer candidates to the question;

executing syntactic analysis processing or syntactic and semantic analysis processing on a passage obtained as a result of the search processing;

generating a tree structure on a basis of a result of the analysis;

extracting a relation between a search word applied in the search processing and each acquired answer candidate as a feature corresponding to each answer candidate from the tree structure generated;

determining an evaluation value of each answer candidate on a basis of the feature extracted; and

re-ranking the answer candidates on a basis of the evaluation values determined.