Question answering system, data search method, and computer program

• US 7,805,303 B2
• Filed: 12/13/2005
• Issued: 09/28/2010
• Est. Priority Date: 04/13/2005
• Status: Expired due to Fees

• Current Assignee: Fujifilm Business Innovation Corp. (Fujifilm Holdings Corporation)
• Original Assignee: Fuji Xerox Company Limited (Fujifilm Holdings Corporation)
• Inventors: Ohkuma, Tomoko, Masuichi, Hiroshi, Yoshimura, Hiroki, Sugihara, Daigo
• Primary Examiner(s): Opsasnick; Michael N
• Application Number:
  US11/299,746
• Publication Number:
  US 20060235689A1
• Time in Patent Office: 1,750 Days
• Field of Search: None
• US Class Current: 704/257
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding

Abstract

A question answering system includes a question input unit, a search unit, an answer candidate extraction unit, an answer candidate inspection unit and an answer output unit. The search unit executes search processing based on an input question. The answer candidate extraction unit extracts an initial answer candidate based on a result of the search processing. The answer candidate inspection unit inspects the initial answer candidate. The answer candidate inspection unit executes search processing with using a query including the initial answer candidate. The answer candidate inspection unit determines whether each word of a sentence obtained as a result of the search processing has a similar lexical meaning to that of each word of the input question. The answer candidate inspection unit selects an initial answer candidate contained in a query determined to have the similar lexical meaning, as a secondary answer candidate.

Claims

1. A question answering system comprising:

a question input unit that receives an input question;

a search unit that executes a first search processing on a basis of the input question;

an answer candidate extraction unit that extracts an initial answer candidate on a basis of a result of the first search processing executed by the search unit;

an answer candidate inspection unit that inspects the initial answer candidate extracted by the answer candidate extraction unit; and

an answer output unit that outputs the initial answer candidate selected by the answer candidate inspection unit, as an answer of the input question, wherein;

the answer candidate inspection unit generates one or more queries and executes a second search processing using the queries, each of the generated queries being a character string pattern relative to the input question and being generated based on one or more keywords included in the input question and one of the initial answer candidates extracted by the answer candidate extraction unit,

the answer candidate inspection unit determines whether or not a composed word of a sentence, which is obtained as a result of the second search processing executed by the answer candidate inspection unit, has a similar lexical meaning to a lexical meaning of a specific word of the input question, and

the answer candidate inspection unit selects the initial answer candidate included in the query, which is used when searching the sentence including the composed word that has the similar lexical meaning, as the answer.

2. The question answering system according to claim 1, wherein

the answer candidate inspection unit executes the second search processing using the query in order to search for a sentence having the same word as the specific word of the input question, and

the answer candidate inspection unit determines whether or not the composed word of the sentence, which has the same word as the specific word of the input question, has the similar lexical meaning to the lexical meaning of the specific word of the input question.

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

a question meaning analysis unit that extracts a question focus from the input question, wherein;

the answer candidate inspection unit generates the query including the question focus extracted by the question meaning analysis unit and the initial answer candidate extracted by the answer candidate extraction unit.

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

a question meaning analysis unit extracts a question focus and a modifier modifying the question focus from the input question, and analyzes a lexical meaning of the modifier as the lexical meaning of the specific word of the input question.

5. The question answering system according to claim 4, wherein the answer candidate inspection unit compares the lexical meaning of the modifier, which is analyzed by the question meaning analysis unit, with the composed word of the sentence obtained as the result of the second search processing.

6. The question answering system according to claim 4, wherein:

the question meaning analysis unit analyzes a semantic distance between the question focus and the modifier in a thesaurus in the analyzing of the lexical meaning of the modifier when existing a plurality of lexical meanings of a single modifier, and

the question meaning analysis unit preferentially selects at least one of the plurality of lexical meanings of the analyzed modifier as the lexical meaning of that modifier, the one of the plurality of lexical meanings having the semantic distance closer to that question focus in the meaning distance.

7. The question answering system according to claim 4, wherein:

the question meaning analysis unit executes syntactic and semantic analysis processing of the input question to extract the question focus and the modifier from the input question.

8. The question answering system according to claim 7, wherein the question meaning analysis unit extracts the question focus and a phrase portion containing a verb directly modifying the question focus from the input question using a case frame dictionary in the syntactic and semantic analysis processing of the input question.

9. A data search method comprising:

inputting an input question;

executing a first search processing on a basis of the input question;

extracting an initial answer candidate on a basis of a result of the first search processing;

inspecting the extracted initial answer candidate; and

outputting the extracted initial answer candidate selected by the inspecting of the initial answer candidate as an answer of the input question, wherein;

the inspecting of the initial answer candidate comprises;

generating one or more queries and executing a second search processing using the queries, each of the generated queries being a character string pattern relative to the input question and being generated based on one or more keywords including in the input question and one of the extracted initial answer candidates;

determining whether or not a composed word of a sentence, which is obtained as a result of the second search processing executed by using the query including the initial answer candidate, has a similar lexical meaning to a lexical meaning of a specific word of the input question; and

selecting the initial answer candidate included in the query, which is used when searching the sentence including the composed word that has the similar lexical meaning, as the answer.

10. The data search method according to claim 9, wherein

the inspecting of the initial answer candidate further comprises:

executing the second search processing using the query in order to search for a sentence having the same word as a specific word of the input question; and

determining whether or not the composed word of the sentence, which has the same word as the specific word of the input question, has the similar lexical meaning to the lexical meaning of the specific word of the input question.

11. The data search method according to claim 9, further comprising:

extracting a question focus from the input question, wherein;

the inspecting of the initial answer candidate comprises;

generating the query including the extracted question focus and the extracted initial answer candidate.

12. The data search method according to claim 9, further comprising:

extracting a question focus and a modifier modifying the question focus from the input question; and

analyzing a lexical meaning of the modifier as the lexical meaning of the specific word of the input question.

13. The data search method according to claim 12, wherein the inspecting of the initial answer candidate further comprising comparing the lexical meaning of the modifier, which is analyzed, with the composed word of the sentence obtained as the result of the second search processing.

14. The data search method according to claim 12, wherein the analyzing of the lexical meaning of the modifier comprises:

analyzing a semantic distance between the question focus and the modifier in a thesaurus when existing a plurality of lexical meanings of a single modifier; and

preferentially selecting at least one of the plurality of lexical meanings of the analyzed modifier as the lexical meaning of that modifier, the one of the plurality of lexical meanings having the semantic distance closer to that question focus in the meaning distance.

15. The data search method according to claim 12 wherein the analyzing of the lexical meaning of the modifier executes syntactic and semantic analysis processing of the input question to extract the question focus and the modifier from the input question.

16. The data search method according to claim 15, wherein the analyzing of the lexical meaning of the modifier further extracts the question focus and a phrase portion containing a verb directly modifying the question focus from the input question using a case frame dictionary in the syntactic and semantic analysis processing of the input question.

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

inputting an input question;

executing a first search processing on a basis of the input question;

extracting an initial answer candidate on a basis of a result of the first search processing;

inspecting the extracted initial answer candidate; and

outputting the extracted initial answer candidate selected by the inspecting of the initial answer candidate as an answer of the input question, wherein;

the inspecting of the initial answer candidate comprises;

generating one or more queries and executing a second search processing using the queries, each of the generated queries being a character string pattern relative to the input question and being generated based on one or more keywords including in the input question and one of the extracted initial answer candidates;

determining whether or not a composed word of a sentence, which is obtained as a result of the second search processing executed by using the query including the initial answer candidate, has a similar lexical meaning to a lexical meaning of a specific word of the input question; and

selecting the initial answer candidate included in the query, which is used when searching the sentence including the composed word that has the similar lexical meaning, as the answer.

18. The computer program stored in a recording medium according to claim 17, wherein

the inspecting of the initial answer candidate further comprises:

executing the second search processing using the query in order to search for a sentence having the same word as a specific word of the input question; and

determining whether or not the composed word of the sentence, which has the same word as the specific word of the input question, has the similar lexical meaning to the lexical meaning of the specific word of the input question.