System and method for generating semantic analysis of textual information

• US 5,983,170 A
• Filed: 06/25/1996
• Issued: 11/09/1999
• Est. Priority Date: 06/25/1996
• Status: Expired due to Fees

• Current Assignee: TopicalNet, Inc. (TransPerfect Global Incorporated)
• Original Assignee: TopicalNet, Inc. (TransPerfect Global Incorporated)
• Inventors: Goodman, Marc I.
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Edouard, Patrick N.
• Application Number:
  US08/670,809
• Time in Patent Office: 1,232 Days
• Field of Search: 704/1, 704/4, 704/9, 704/8, 364/975
• US Class Current: 704/9
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A system receives an information stream comprising the textual information whose semantic content is to be determined, divides the information stream into a series of elements and classifies each element into one of a plurality of predetermined classifications. The system includes a token generator and a token classifier. The token generator receives the textual information stream, parses the stream to identifies the respective elements, identifies for each element one of a plurality of element types, and generates a token identifying the element type for each element. At least some of the tokens also include a pointer pointing to the actual information associated with the element. The token classifier receives the tokens and classifies them in order. In that operation, the token classifier classifies each token in relation to the token'"'"'s type, classifications for previously-classified tokens and the types of successive tokens, thereby to determine the semantic content of the information associated with the tokens. After the tokens are classified, the information associated therewith can be loaded into a database system according to their classifications, and conventional database tools used to obtain information therefrom.

Claims

1. A system for receiving an information stream in the form of textual information comprising a series of information elements and generating therefrom respective classifications for said information elements from a plurality of predetermined classifications, said system comprising:

A. a token generator configured to receive the information stream, parse the information stream to identify the respective information elements, identify for each information element one of a plurality of element types, and generate for each information element a token identifying the information element'"'"'s element type;

B. a token classifier configured to receive the tokens and generate a classification to classify each said token in relation to the element type associated with said respective token, classifications generated for previously-classified tokens and the types of previous and successive tokens, thereby to determine the semantic content of the information associated with the tokens.

2. A system as defined in claim 1 further comprising a database system, the information elements being loaded into the database system according to their respective classifications.

3. A system as defined in claim 1 in which the token classifier includes a trained classification tree.

4. A system as defined in claim 3 in which the trained classification tree is trained using a training set of properly classified tokens.

5. A system as defined in claim 4 further comprising a maintenance module for retraining the classification tree if it determines that a token is incorrectly classified, using the training set expanded by the incorrectly classified token, for which an operator has provided a proper classification.

6. A method of receiving an information stream in the form of textual information comprising a series of information elements and generating therefrom respective classifications for said information elements from a plurality of predetermined classifications, said method comprising the steps of:

A. receiving the information stream, parsing the information stream to identify the respective information elements, identifying for each information element one of a plurality of element types, and generating for each information element a token identifying the information element'"'"'s element type;

B. generating for each token a classification to classify each said token in relation to the element type associated with said respective token, classifications generated for previously-classified tokens and the types of previous and successive tokens, thereby to determine the semantic content of the information associated with the tokens.

7. A method as defined in claim 6 further comprising the step of loading the tokens into a database system according to their respective classifications.

8. A method as defined in claim 6 in which the token classification step includes use of a trained classification tree.

9. A method as defined in claim 8 includes the step of training the classification tree using a training set of properly classified tokens.

10. A method as defined in claim 9 further comprising the step of retraining the classification tree if the classification tree determines that a token is incorrectly classified, using the training set expanded by the incorrectly classified token, for which an operator has provided a proper classification.

11. A computer program product for controlling a computer to receiving an information stream in the form of textual information comprising a series of information elements and generating therefrom respective classifications for said information elements from a plurality of predetermined classifications, said computer program product comprising a machine-readable medium having encoded thereon:

A. a token generator module configured to enable the computer to receive the information stream, parse the information stream to identify the respective information elements, identify for each information element one of a plurality of element types, and generate for each information element a token identifying the information element'"'"'s element type;

B. a token classifier module configured to enable the computer to, for each token, generate a classification to classify each said token in relation to the element type associated with said respective token, classifications generated for previously-classified tokens and the types of previous and successive tokens, thereby to determine the semantic content of the information associated with the tokens.

12. A computer program product as defined in claim 11 further comprising a database system module for controlling the computer to load the information elements into a database according to their respective classifications.

13. A computer program product as defined in claim 11 in which the token classifier module includes a trained classification tree.

14. A computer program product as defined in claim 13 in which the trained classification tree is trained using a training set of properly classified tokens.

15. A computer program product as defined in claim 14 further comprising a maintenance module for enabling the computer to retrain the classification tree if it determines that a token is incorrectly classified, using the training set expanded by the incorrectly classified token, for which an operator has provided a proper classification.