INFERRING LEXICAL ANSWER TYPES OF QUESTIONS FROM CONTEXT

• US 20110078192A1
• Filed: 09/30/2009
• Published: 03/31/2011
• Est. Priority Date: 09/30/2009
• Status: Abandoned Application

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Murdock, James W. IV
• Application Number:
  US12/570,575
• Publication Number:
  US 20110078192A1
• Time in Patent Office: Days
• US Class Current: 707/780
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems

Abstract

A method, system and computer program product are disclosed for searching for an inferred lexical answer type from the context of a question. One embodiment comprises determining from a question an original lexical answer type (LAT), extracting the context in which said original LAT occurs within the question, and identifying a set of context fragments from said context. This embodiment further comprises using the set of context fragments to search through a defined database to produce a search result including a plurality of text strings having a specified relationship with said context fragments, and processing said search result to search for one or more inferred LAT of the question. In one embodiment, the processing includes extracting a plurality of LATs from the text strings in the search result, and selecting one or more of the extracted LATs as the one or more inferred LAT of the question.

Claims

1. A method of searching for an inferred lexical answer type from the context of a question, comprising:

receiving a question;

determining from the question an original lexical answer type (LAT);

extracting the context in which said original LAT occurs within the question;

identifying a set of context fragments from said context;

using the set of context fragments in a given procedure for searching through a defined database to produce a search result including a plurality of text strings having a specified relationship with said context fragments; and

processing said search result to search for one or more inferred LAT of the question.

2. The method according to claim 1, wherein the processing includes:

extracting a plurality of LATs from the text strings in the search results; and

selecting one or more of the extracted LATs.

3. The method according to claim 2, wherein the selecting includes:

ranking said extracted LATs according to a given criteria; and

selecting said one of the extracted LATs on the basis of said ranking.

4. The method according to claim 3, wherein said ranking is based on the frequency with which each of the extracted LATs occurs in said plurality of text strings.

5. The method according to claim 2, wherein the extracting a plurality of LATs from the text strings includes, for each of the plurality of text strings:

aligning one of the context fragments including the original LAT with said each of the text strings; and

extracting the term in said each of the text strings that corresponds with the original LAT in said one of the context fragments.

6. The method according to claim 1, wherein the using the set of context fragments in the given procedure includes:

producing abstract versions of the context fragments;

inputting the context fragments and the abstract versions of the context fragments to a search component;

said search component searching through the defined database for text strings that match either the context fragments or the abstract versions of the context fragments according to defined criteria.

7. The method according to claim 6, wherein the producing abstract versions of the context fragments includes replacing selected words in the context fragments with other words that are more general than said selected words.

8. The method according to claim 6, wherein the producing abstract versions of the context fragments includes replacing selected words in the context fragments with terms in structured ontologies.

9. The method according to claim 1, wherein the context fragments are determined by lexical ordering.

10. The method according to claim 1, wherein the context fragments are determined by semantic relationships.

11. A system for searching for an inferred lexical answer type from the context of a question, the system comprising one or more processing units configured for:

receiving a question;

determining from the question an original lexical answer type (LAT);

extracting the context in which said original LAT occurs within the question;

identifying a set of context fragments from said context;

using the set of context fragments in a given procedure for searching through a defined database to produce a search result including a plurality of text strings having a specified relationship with said context fragments; and

processing said search result to search for one or more inferred LAT of the question.

12. The system according to claim 11, wherein the processing includes:

extracting a plurality of LATs from the text strings in the search results; and

selecting one or more of the extracted LATs.

13. The system according to claim 12, wherein the selecting one of the extracted LATs includes:

ranking said extracted LATs according to a given criteria; and

selecting said one of the extracted LATs on the basis of said ranking; and

wherein said ranking is based on the frequency with which each of the extracted LATs occurs in said plurality of text strings.

14. The system according to claim 12, wherein the extracting a plurality of LATs from the text strings includes, for each of the plurality of text strings:

aligning one of the context fragments including the original LAT with said each of the text strings; and

extracting the term in said each of the text strings that corresponds with the original LAT in said one of the context fragments.

15. The system according to claim 11, wherein the using the set of context fragments in the given procedure includes:

producing abstract versions of the context fragments;

inputting the context fragments and the abstract versions of the context fragments to a search component;

said search component searching through the defined database for text strings that match either the context fragments or the abstract versions of the context fragments according to defined criteria.

16. An article of manufacture comprising:

at least one computer usable medium having computer readable program code logic to execute instructions in a processing unit for searching for an inferred lexical answer type from the context of a question, said computer readable program code logic, when executing, performing the following;

receiving a question;

determining from the question an original lexical answer type (LAT);

extracting the context in which said original LAT occurs within the question;

identifying a set of context fragments from said context;

using the set of context fragments in a given procedure for searching through a defined database to produce a search result including a plurality of text strings having a specified relationship with said context fragments; and

processing said search result to search for one or more inferred LAT of the question.

17. The article of manufacture according to claim 16, wherein the processing includes:

extracting a plurality of LATs from the text strings in the search results; and

selecting one or more of the extracted LATs.

18. The article of manufacture according to claim 17, wherein the selecting one of the extracted LATs includes:

ranking said extracted LATs according to a given criteria; and

selecting said one of the extracted LATs on the basis of said ranking; and

wherein;

said ranking is based on the frequency with which each of the extracted LATs occurs in said plurality of text strings.

19. The article of manufacture according to claim 17, wherein the extracting a plurality of LATs from the text strings includes, for each of the plurality of text strings:

aligning one of the context fragments including the original LAT with said each of the text strings; and

extracting the term in said each of the text strings that corresponds with the original LAT in said one of the context fragments.

20. The article of manufacture according to claim 16, wherein the using the set of context fragments in the given procedure includes:

producing abstract versions of the context fragments;

inputting the context fragments and the abstract versions of the context fragments to a search component;

said search component searching through the defined database for text strings that match either the context fragments or the abstract versions of the context fragments according to defined criteria.