Information Extraction System

• US 20070255553A1
• Filed: 03/31/2005
• Published: 11/01/2007
• Est. Priority Date: 03/31/2004
• Status: Abandoned Application

• Current Assignee: Panasonic Corporation (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Nomoto, Masako
• Application Number:
  US10/599,507
• Publication Number:
  US 20070255553A1
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

Contents of a description such as a fact and an opinion related to a thing expressed in a text are extracted with a viewpoint of the fact and the opinion related to a description. To execute it, there are provided an input unit that inputs a text, a viewpoint and description extraction rule storage that stores a viewpoint and description extraction rule for specifying a pair of a viewpoint of an expression described in the text and a description of the viewpoint, a viewpoint and description extraction unit that relates the pair of the viewpoint and its description based upon at least one of a syntactic attribute and a semantic attribute of a character string in the text using the viewpoint and description extraction rule and extracts the pair as element metadata including identification information for identifying the pair, and a metadata storage that stores the element metadata extracted by the viewpoint and description extraction unit. The contents of a description such as a fact and an opinion related to a thing expressed in the text are configured as a pair of a viewpoint and a description, are arranged in a form in which the fact and the opinion are related, and are extracted.

Claims

1. An information extraction system, comprising:

an input unit that inputs a text;

a viewpoint and description extraction rule storage that stores a viewpoint and description extraction rule for specifying the pairs of a viewpoint and a description for an expression in the text;

a viewpoint and description extraction unit that extracts the corresponding pairs of the viewpoint and its description of the expression using the viewpoint and description extraction rules that based on syntactic and/or semantic attributes and extract them as the element metadata to which identification information is added; and

a metadata storage that stores element metadata extracted by the viewpoint and description extraction unit.

2. The information extraction system according to claim 1,

wherein said syntactic attribute includes the character string notation and/or the classification of a part of speech.

3. The information extraction system according to claim 1,

wherein the semantic attribute includes semantic classification.

4. The information extraction system according to claim 1,

wherein the viewpoint and description extraction unit extracts the corresponding pair of the viewpoint and its description of the expression using the viewpoint and description as the element metadata to which identification information is added.

5. The information extraction system according to claim 1, further comprising:

an attribute addition unit that adds the semantic attribute including the semantic classification for character strings in a text using the semantic attribute addition rules and outputs the text added the semantic attributes.

6. The information extraction system according to claim 1,

wherein the viewpoint and description extraction unit extracts the corresponding pair of the viewpoint and its description using a semantic attribute of the description for the viewpoint when no viewpoint is explicitly expressed in a text and only the description is expressed.

7. The information extraction system according to claim 1, further comprising:

a metadata comparison unit that compares and estimates the respective relationships between the viewpoints and between the descriptions of element metadata extracted by the viewpoint and description extraction unit; and

a metadata integration unit that integrates related element metadata based upon the estimated relevance and outputs integrated metadata.

8. The information extraction system according to claim 6,

wherein the metadata comparison unit compares respective relationships between the viewpoints and between the descriptions of element metadata extracted by the viewpoint and description extraction unit using at least the semantic attributes of the character strings comprising the viewpoint and or description.

9. The information extraction system according to claim 7, further comprising:

a topical thing estimation unit that estimates a topical thing in element metadata extracted by the viewpoint and description extraction unit using a topical thing estimation rule for estimating a topical thing,

wherein the metadata storage also relates and stores the topical thing estimated by the topical thing estimated by the topical thing estimation unit together with the corresponding element metadata.

10. The information extraction system according to claim 9,

wherein the topical thing estimation unit estimates a topical thing of an element metadata based upon a viewpoints, descriptions, and semantic attributes of the element metadata stored in the metadata storage.

11. The information extraction system according to claim 9 or 10,

wherein the metadata comparison unit compares between viewpoints and between descriptions of the same topical thing estimated by the topical thing estimation unit.

12. The information extraction system according to claim 1,

wherein the viewpoint and description extraction rule includes a user viewpoint and description extraction rule which is a rule for specifying information on user who creates the text;

and

the viewpoint and description extraction unit extracts user metadata which is element metadata related to the user information using the user viewpoint and description extraction rule.

13. The information extraction system according to claim 1,

wherein the viewpoint and description extraction rule includes a source viewpoint and description extraction rule which is a rule for specifying source information that describes the bibliography of the text;

and

the viewpoint and description extraction unit extracts source metadata which is element metadata related to the source information using the source viewpoint and description extraction rule.

14. The information extraction system according to claim 9,

wherein the metadata comparison unit is further provided with an objectivity/reliability determination unit that determines objectivity and reliability of a viewpoint and a description using at least one of element metadata, user metadata and source metadata, and an objectivity/reliability determination rule storage that stores an objectivity/reliability determination rule for determining the objectivity and the reliability of the viewpoint and the description.

15. The information extraction system according to claim 9,

wherein when the topical thing estimation unit estimates a topical thing of element metadata, it estimates the topical thing using at least one of source metadata and user metadata in addition to the element metadata stored in the metadata storage.

16. The information extraction system according to claim 14, further comprising:

a metadata output format generator that arranges the metadata in a format of a table and generates a metadata table; and

a metadata output unit that shows the generated metadata table to a user.

17. The information extraction system according to claim 16, further comprising:

a user request processor that processes a request from a user,

wherein the metadata output format generator generates a metadata table using metadata matched with the request from the user input via the user request processor.

18. An information extraction method, comprising:

a step for inputting a text;

a step for referring to a viewpoint and description extraction rule for specifying a pair of a viewpoint of an expression described in the text and a description of the viewpoint; and

a step for extracting the pair of the viewpoint and its description based upon at least one of a syntactic attribute and a semantic attribute added to a character string in the text input to an input unit using the viewpoint and description extraction rule as element metadata relating the pair.