Question answering system, data search method, and computer program

• US 7,587,389 B2
• Filed: 12/12/2005
• Issued: 09/08/2009
• Est. Priority Date: 06/10/2005
• Status: Expired due to Fees

• Current Assignee: Fuji Xerox Company Limited (Xerox Holdings Corp.)
• Original Assignee: Fuji Xerox Company Limited (Xerox Holdings Corp.)
• Inventors: Ohkuma, Tomoko, Masuichi, Hiroshi, Yoshimura, Hiroki, Sugihara, Daigo
• Primary Examiner(s): Vy; Hung T
• Application Number:
  US11/298,855
• Publication Number:
  US 20060282414A1
• Time in Patent Office: 1,366 Days
• Field of Search: 704/1, 704 4- 5, 704/231, 704/9, 707 3- 5
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching

Abstract

A question answering system includes a question input unit input to a question, a search unit, an anaphoric analysis unit, an information extraction unit and an answer candidate extraction unit. The search unit executes search processing based on the input question. The anaphoric analysis unit executes anaphoric analysis processing on a search result. The information extraction unit extracts sentences, which include expressions having a reference relation therebetween, as reference relation sentences, the reference relation determined by finding relation in an expressions of at least one of a noun phrase, a pronoun, and a zero pronoun between sentences in the search result on a basis of a result of the anaphoric analysis; selects a first candidate sentence from a passage obtained as the search result; and selects a second candidate sentence, which is the reference relation sentence of the first candidate sentence. The answer candidate extraction unit extracts an answer candidate.

Claims

1. A question answering system comprising:

a CPU;

a question input unit that inputs a question;

a search unit that executes search processing based on the input question;

an anaphoric analysis unit that executes anaphoric analysis processing on a search result of the search unit;

an information extraction unit that extracts sentences, which include expressions having a reference relation therebetween, as reference relation sentences, the reference relation determined by finding an anaphoric relation between a first expression and a second expression of at least one of a noun phrase, a pronoun, and a zero pronoun between respective sentences in the search result based on a result of the anaphoric analysis executed by the anaphoric analysis unit, wherein

the information extraction unit selects a first candidate sentence containing the first expression from a passage obtained as the search result, and selects a second candidate sentence containing the second expression, which is a reference relation sentence of the first candidate sentence; and

the information extraction unit selects as the first candidate sentence from the passage obtained as the search result, a sentence containing more words corresponding to a search keyword used by the search unit and another sentence adjacent thereto; and

if a reference relation sentence of the selected first candidate sentence is present, the information extraction unit selects the reference relation sentence of the selected first candidate sentence as the second candidate sentence; and

an answer candidate extraction unit that extracts an answer candidate from the first and second candidate sentences selected by the information extraction unit.

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

an answer selection unit that calculates a score of the extracted answer candidate indicating suitability degree as an answer to the input question.

3. The question answering system according to claim 2, wherein:

the answer selection unit calculates a first score value of the answer candidate on a basis of number of words, which correspond to the search word and are contained in at least one of a sentence containing the answer candidate and a sentence adjacent to the sentence containing the answer candidate,

the answer selection unit calculates a second score value of the answer candidate on a basis of number of words, which correspond to the search word and are contained in at least one of a sentence containing the answer candidate and a reference relation sentence of a sentence adjacent to the sentence containing the answer candidate, and

the answer selection unit calculates the score of the answer candidate on a basis of the first and second score values of the answer candidate.

4. A data search method comprising:

inputting a question;

executing search processing based on the input question;

executing anaphoric analysis processing on a search result;

extracting sentences, which include expressions having a reference relation therebetween, as reference relation sentences, the reference relation determined by finding an anaphoric relation between a first expression and a second expression of at least one of a noun phrase, a pronoun, and a zero pronoun between respective sentences in the search result based on a result of the anaphoric analysis;

selecting a first candidate sentence containing the first expression from a passage obtained as the search result;

selecting a second candidate sentence containing the second expression, which is the reference relation sentence of the first candidate sentence containing the first expression; and

extracting an answer candidate from the first and second candidate sentences selected, wherein

the selecting of the first candidate sentence comprises selecting as the first candidate sentence from the passage obtained as the search result, a sentence containing more words corresponding to a search keyword used by the search unit and another sentence adjacent thereto,

the selecting of the second candidate sentence comprises, if a reference relation sentence of the selected first candidate sentence is present, selecting the reference relation sentence of the selected first candidate sentence as the second candidate sentence, and

a processor performs the method.

5. The data search method according to claim 4, further comprising:

calculating a score of the extracted answer candidate indicating suitability degree as an answer to the input question.

6. The data search method according to claim 5, wherein

the calculating of the score comprises:

calculating a first score value of the answer candidate on a basis of number of words, which correspond to the search word and are contained in at least one of a sentence containing the answer candidate and a sentence adjacent to the sentence containing the answer candidate;

calculating a second score value of the answer candidate on a basis of number of words, which correspond to the search word and are contained in at least one of a sentence containing the answer candidate and a reference relation sentence of a sentence adjacent to the sentence containing the answer candidate; and

calculating the score of the answer candidate on a basis of the first and second score values of the answer candidate.

7. A computer-readable medium storing a computer program causing a computer to execute data search processing, the data search processing comprising:

inputting a question;

executing search processing based on the input question;

executing anaphoric analysis processing on a search result;

extracting sentences, which include expressions having a reference relation therebetween, as reference relation sentences, the reference relation determined by finding an anaphoric relation between a first expression and a second expression of at least one of a noun phrase, a pronoun, and a zero pronoun between respective sentences in the search result on a basis of a result of the anaphoric analysis;

selecting a first candidate sentence containing the first expression from a passage obtained as the search result;

selecting a second candidate sentence containing the second expression, which is the reference relation sentence of the first candidate sentence containing the first expression; and

extracting an answer candidate from the first and second candidate sentences wherein

the selecting of the first candidate sentences comprises selecting as the first candidate sentence from the passage obtained as the search result, a sentence containing more words corresponding to a search keyword used by the search unit and another sentence adjacent thereto, and

the selecting of the second candidate sentence comprises, if a reference relation sentence of the selected first candidate sentence is present, selecting the reference relation sentence of the selected first candidate sentence as the second candidate sentence.