System and method for building a semantic network capable of identifying word patterns in text

• US 6,813,616 B2
• Filed: 03/07/2001
• Issued: 11/02/2004
• Est. Priority Date: 03/07/2001
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Simpson, Don M., Usey, Robert W. Jr.
• Primary Examiner(s): Amsbury, Wayne
• Assistant Examiner(s): NGUYEN, CAM LINH T
• Application Number:
  US09/800,625
• Publication Number:
  US 20030167162A1
• Time in Patent Office: 1,336 Days
• Field of Search: 707/3, 707/102
• US Class Current: 1/1
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

A semantic network is disclosed herein that is useful for identifying word patterns in text. The semantic network may be used to identify word patterns accurately and substantially in real time. The semantic network is comprised of recognized words, which form word patterns, and objects, which are identified by certain of the word patterns. The word patterns are located at nodes in the semantic network. The nodes are linked in a hierarchical structure, with varying, identifiable levels. Under the method of building a semantic network, a computer program is provided and loaded onto a computer. The software is preferably linked with a knowledge base from which identified objects can be obtained. The objects are each classified by object type as they are received. Each object has a label which corresponds to a upper level word pattern. The word pattern is parsed or “tokenized” and combinations of the words in the label are assigned as word patterns at various nodes. The word patterns that together form a higher level word pattern are linked together and word patterns are also labeled according to their level in the hierarchy of the semantic network. The word patterns corresponding to labels of objects are linked to the objects they identify.

Claims

1. A method for generating a semantic network capable of identifying word patterns in text and associating the word patterns with known objects, the method comprising:

receiving a plurality of objects from a data source;

tokenizing labels of the objects into separate words, the labels uniquely identifying the objects;

building word patterns from the separate words of each label, each word pattern comprising a node in a semantic network comprising word patterns and the objects;

linking the word patterns to other word patterns at different hierarchical levels within the semantic network; and

associating selected word patterns with objects identified by the word patterns.

2. The method of claim 1, wherein the step of receiving objects from a data source further comprises placing the objects into nodes in the semantic network and linking the objects to a data source.

3. The method of claim 2, wherein the data source comprises a knowledge base.

4. The method of claim 1, further comprising placing each separate word into a node of the semantic network.

5. The method of claim 1, further comprising linking each word with corresponding word patterns comprising a group of words.

6. The method of claim 1, wherein building word patterns further comprises assigning each word pattern a hierarchical level in the semantic network and identifying the hierarchical level of each word pattern.

7. The method of claim 6, wherein the hierarchical levels vary according to the number of words in a word pattern.

8. The method of claim 1, wherein the steps of claim 1 take place within a common RAM of a computer.

9. The method of claim 1, further comprising associating the word pattern comprising the label of an object with that object.

10. The method of claim 1, further comprising identifying types of the objects that are to be loaded into the semantic network, and assigning one of the identified object types to each object.

11. A system for generating a semantic network capable of identifying word patterns in text and associating the word patterns with known objects, the system comprising:

a data source interface configured to receive a plurality of objects;

an object parser module configured to parse labels of the objects into separate words, the labels uniquely identifying the objects;

a word pattern placement module configured to build word patterns from the separate words of each label, each word pattern comprising a node in a semantic network comprising word patterns and the objects;

a node linking module configured to link the word patterns to other word patterns at different hierarchical levels within the semantic network; and

an object association module configured to associate selected word patterns with objects identified by the word patterns.

12. The system of claim 11, further comprising a data source and wherein the plurality of objects are received from the data source.

13. The system of claim 12, wherein the data source comprises a knowledge base.

14. The system of claim 11, further comprising a word placement module configured to place each separate word into a node of the semantic network.

15. The system of claim 11, wherein the node linking module is further configured to link each word with corresponding word patterns comprising a group of words of which the word is a part.

16. The system of claim 11, wherein the word pattern placement module is further configured to assigning each word pattern a hierarchical level in the semantic network and to identify the hierarchical level of each word pattern.

17. The system of claim 16, wherein the hierarchical levels vary according to the number of words in a word pattern.

18. The system of claim 11, wherein the system is configured to operate within RAM memory of a computer.

19. The system of claim 11, further comprising an object association module configured to associate a word pattern comprising the label of an object with that object.

20. The system of claim 11, further comprising an object labeling module configured to identify types of the objects that are to be loaded into the semantic network and to assign one of the identified object types to each object.