Topic map based indexing and searching apparatus

• US 8,554,540 B2
• Filed: 06/15/2009
• Issued: 10/08/2013
• Est. Priority Date: 12/11/2008
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: Lee, Chung Hee, Oh, Hyo-Jung, Heo, Jeong, Hwang, Yi Gyu, Yoon, Yeo Chan, Choi, Miran, Lee, Chang Ki, Lim, Soojong, Kim, HyunKi, Jang, Myung Gil
• Primary Examiner(s): COLUCCI, MICHAEL C
• Application Number:
  US12/484,651
• Publication Number:
  US 20100153094A1
• Time in Patent Office: 1,576 Days
• Field of Search: 704/9, 706/42, 706/50, 706/47, 706/11, 717/126, 717/104, 707/608, 705/7.17, 703/27
• US Class Current: 704/9
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems

Abstract

A topic map based indexing apparatus analyzes community Q/A lists to acquire Q/A analysis information, removes redundant answers depending on the Q/A analysis information, removes insignificant answers based on the degree of reliability, ranks answer lists, and extracts the highest ranking answer as a best answer, to thereby store, in a community Q/A topic map, index information containing the community Q/A lists and the Q/A analysis information. A topic map based searching apparatus analyzes a user question to acquire question analysis information, searches similar questions from community Q/A lists belonging to a specific topic node of a pre-stored community Q/A topic map, ranks the searched similar questions depending on the question analysis information, removes redundant answers among answers to the ranked similar questions, ranks the answers, and extracts the highest ranking answer as a best answer.

Claims

1. A topic map based indexing apparatus, comprising:

a question/answer (Q/A) pre-processing block to, by a computer, normalize community Q/A lists for plural different community sites on web, respectively, correct grammatical errors, and remove insignificant symbols, to thereby provide pre-processed community Q/A lists for the plural different community sites, respectively;

a Q/A analysis block to, by a computer, analyze the pre-processed community Q/A lists to acquire Q/A analysis information; and

a Q/A storage block to, by a computer, store, in a community Q/A topic map, index information containing the pre-processed community Q/A lists and the Q/A analysis information, wherein the indexing information is obtained by removing redundant answers depending on the Q/A analysis information, removing insignificant answers based on a degree of reliability, ranking answer lists, extracting a highest ranking answer as a best answer, and determining a topic,

wherein the Q/A pre-processing block includes;

a one unit recognizing unit to normalize, in a single form, allomorph words or words recognizable as one unit in the community Q/A lists by referring to a unit dictionary database; and

an error pre-processing unit to correct the grammatical errors and remove the insignificant symbols from the community Q/A lists by referring to a pre-processing rule database.

2. The topic map based indexing apparatus of claim 1, wherein the Q/A analysis block includes:

a language analysis unit to perform language analysis on the pre-processed community Q/A lists by morpheme analysis, named entity recognition, and partial syntactic analysis, to thereby provide language analysis results;

a domain classification unit to determine domains of the pre-processed community Q/A lists by referring to a domain classification database;

a topic classification unit to classify the pre-processed community Q/A lists by topic by referring to a topic classification database;

a Q/A type analysis unit to analyze a type of an expected community answer in the pre-processed community Q/A lists based on an intention of a community question with reference to a Q/A type database;

a keyword extraction and extension unit to extract keywords from the pre-processed community Q/A lists depending on the language analysis results by referring to a keyword extension dictionary database, and to additionally extend a range of similar words corresponding to the keywords; and

a constraint word extraction unit to extract, from the pre-processed community Q/A lists, constraint words in order to restrict a search space by referring to a constraint word extraction rule database.

3. The topic map based indexing apparatus of claim 1, wherein the Q/A storage block includes:

a redundant answer removal unit to remove redundant answers having the same meaning from the pre-processed community Q/A lists to a community question to leave only one answer;

a wrong answer filtering unit to remove insignificant answers from the pre-processed community Q/A lists provided from the redundant answer removal unit by referring to a filtering rule database based on the degree of reliability;

a best answer extraction unit to rank answers of the pre-processed community Q/A lists provided from the wrong answer filtering unit, and to extract the highest ranking answer as the best answer by referring to a best answer extraction rule database;

a topic node determination unit to determine a topic appropriate for a community question depending on a classification of topics; and

an index information storage unit to generate the index information containing the community Q/A lists from which the highest ranking best answer is extracted and the Q/A analysis information, and to store the index information in topic nodes corresponding to the determined topic for storage of the community Q/A topic map.

4. The topic map based indexing apparatus of claim 1, wherein the community Q/A topic map has a tree structure with a plurality of topic nodes for navigating to respective pre-processed community Q/A lists, based on the index information and the Q/A analysis information.

5. A topic map based searching apparatus, comprising:

a question pre-processing block to, by a computer, normalize a user question in a single form, correct grammatical errors and remove insignificant symbols, to thereby provide a pre-processed user question;

a question analysis block to, by a computer, analyze the pre-processed user question to acquire question analysis information;

a similar question search block to, by a computer,

limit a search space to a specific topic node of a pre-stored community question/answer (Q/A) topic map depending on the question analysis information, the Q/A topic map having stored therein index information containing pre-processed community Q/A lists for plural different community sites on web, respectively, and Q/A analysis information acquired by analyzing the pre-processed community Q/A lists,

search similar questions from community Q/A lists belonging to the limited search space, and

rank the searched similar questions depending on the question analysis information; and

an answer extraction block to, by a computer, remove redundant answers among answers to the ranked similar questions, rank the answers, and extract a highest ranking answer as a best answer,

wherein the question pre-processing block includes;

a one unit recognizing unit to normalize, in a single form, allomorph words or words recognizable as one unit with respect to the user question by referring to a unit dictionary database; and

an error pre-processing unit to correct the grammatical errors in the user question and remove the insignificant symbols from the user question by referring to a pre-processing rule database.

6. The topic map based searching apparatus of claim 5, wherein the question analysis block includes:

a language analysis unit to perform language analysis on the pre-processed user question by morpheme analysis, named entity recognition, and partial syntactic analysis, to thereby provide a language analysis;

a domain classification unit to determine a domain of the pre-processed user question by referring to a domain classification database;

a topic classification unit to classify the pre-processed user question by topic by referring to a topic classification database;

a Q/A type analysis unit to analyze a type of a desired answer depending on an intention of the user question by referring to a Q/A type database;

a keyword extraction and extension unit to extract keywords from the pre-processed user question on the basis of the language analysis by referring to a keyword extension dictionary database, and to additionally extend a range of similar words corresponding to the keywords; and

a constraint word extraction unit to extract constraint words, which are restrictions in a search of time and location from the pre-processed user question by referring to a constraint word extraction rule database.

7. The topic map based searching apparatus of claim 5, wherein the similar question search block includes:

a search space constraining unit to limit the search space from the community Q/A lists from which an answer to the user question is to be searched to the lists at the specific topic node based on a domain classification and topic classification among the question analysis information;

a semantic-based matching unit to search similar questions of the community Q/A lists at the specific topic node by using the keywords and similar words based on the question analysis information; and

a similar question ranking unit to detect the community Q/A lists containing the searched similar questions and rank the community Q/A lists based on the question analysis information.

8. The topic map based searching apparatus of claim 5, wherein the answer extraction block includes:

a redundant answer removal unit to remove redundant answers with the same meaning among answer lists to the similar questions in the ranked community Q/A lists; and

a best answer ranking unit to rank answers to the similar questions from which the redundant answers have been removed, and provide the highest ranking answer as the best answer.

9. The topic map based searching apparatus of claim 5, wherein the community Q/A topic map has a tree structure with a plurality of topic nodes for navigating to respective pre-processed community Q/A lists, based on the index information and the Q/A analysis information.