Method and device for parsing natural language sentences and other sequential symbolic expressions

• US 5,960,384 A
• Filed: 09/03/1997
• Issued: 09/28/1999
• Est. Priority Date: 09/03/1997
• Status: Expired due to Term

• Current Assignee: Douglas E. Brash
• Original Assignee: Douglas E. Brash
• Inventors: Brash, Douglas E.
• Primary Examiner(s): Thomas, Joseph
• Application Number:
  US08/922,494
• Time in Patent Office: 755 Days
• Field of Search: 704/1, 704/8, 704/9, 704/10, 704/270, 704/271, 704/275, 704/231, 704/235
• US Class Current: 704/9
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A parsing method and apparatus for symbolic expressions of thought such as English-language sentences is provided. The basic procedure distinguishes only between symbols for picture-like entities (such as "squirrel" or "justice") and symbols for relations (such as "above" or "pushed"). For example, to the cognitive parser the sentences "The squirrel buried a nut" and "My aunts sent a telegram" are equivalent. The parser thus operates at a level more basic than syntactic parsing, making it simpler. The illustrative embodiment, a cognitive parser for English sentences, comprises: a microprocessor, a stored lexicon including symbols and associated entity-types (picture or relation), an input device for a symbol sequence, and a procedure executing on the microprocessor for grouping the inputted picture-designating and relation-designating symbols according to rules based on entity type. A method comprising operation of the cognitive parser is also provided.

Claims

1. An apparatus for parsing symbol sequences such as natural language sentences into an organized symbol hierarchy, comprising:

a microprocessor;

a memory addressable by said microprocessor;

a lexicon, stored in said memory, including symbols and an entity-type corresponding to each symbol, the entity-type selected from the group consisting of a picture-like entity, a relation, and a picture-like entity accompanied by a relation;

an input device for inputting the symbol sequence to said microprocessor;

a procedure, executing on said microprocessor, for determining the entity-type corresponding to each inputted symbol by searching said lexicon, and for grouping the inputted symbols according to rules based on entity-type; and

an output device for outputting the symbol grouping.

2. The apparatus of claim 1, wherein said input device comprises a keyboard.

3. The apparatus of claim 1, wherein said input device comprises a voice-recognition system.

4. The apparatus of claim 1, wherein said output device comprises a display screen.

5. The apparatus of claim 1, wherein said output device comprises a software program for using the symbol grouping.

6. The apparatus of claim 5, wherein said software program operates a control unit of another device.

7. The apparatus of claim 1, wherein said output device comprises an information storage device.

8. The apparatus of claim 1, wherein said lexicon includes a cognitive category valuation for each symbol that designates a picture-like entity, the cognitive category substantially ranging from sensation-related to abstract concepts;

and wherein said program includes a consecutive picture-symbol routine that uses the cognitive category valuation for inserting a relation-designating symbol between consecutive symbols which designate picture-like entities.

9. The apparatus of claim 1, wherein said procedure includes routines selected from the group consisting of a picture-symbol grouping routine, a hierarchy-building routine, and a track-utilization routine;

and wherein said procedure uses said routines for building a symbol group consisting of picture-designating symbols and relation-designating symbols.

10. The apparatus of claim 1, wherein said lexicon includes symbol-grouping properties corresponding to at least some of the symbols;

and wherein said procedure also determines the symbol-grouping properties corresponding to an inputted symbol and also groups the inputted picture-designating and relation-designating symbols into a symbol group according to the symbol-grouping properties.

11. A lexicon for use in a parser for symbol sequences such as natural language sentences comprising:

a memory device; and

a table, stored in said memory device, including a plurality of symbols and an entity-type corresponding to each symbol of the plurality of symbols, the entity-type selected from the group consisting of a picture-like entity, a relation, and a picture-like entity accompanied by a relation.

12. The lexicon of claim 11, wherein said table further comprises a cognitive category valuation for each symbol of the plurality of symbols which designates a picture-like entity, the cognitive category valuation substantially ranging from sensation-related to abstract concepts.

13. The lexicon of claim 11, wherein said table further comprises symbol-grouping properties associated with at least one of the plurality of symbols, said symbol-grouping properties applying to picture-designating and relation-designating symbols of said symbol sequences.

14. An apparatus operated by symbol sequences such as natural language sentences, comprising:

a microprocessor;

a memory addressable by said microprocessor;

a lexicon, stored in said memory, including symbols and an entity-type corresponding to each symbol, the entity-type selected from the group consisting of a picture-like entity, a relation, and a picture-like entity accompanied by a relation;

an input device for inputting the symbol sequence to said microprocessor; and

a program, executing on said microprocessor, for determining the entity-type corresponding to each inputted symbol by searching said lexicon, for grouping the inputted symbols according to rules based on entity-type, and for generating an instruction for operating the apparatus in response to the symbol grouping.

15. The apparatus of claim 14, wherein said input device comprises a keyboard.

16. The apparatus of claim 14, wherein said input device comprises a voice-recognition system.

17. A method for parsing symbol sequences such as natural language sentences into an organized symbol hierarchy, comprising the steps of:

receiving a symbol sequence;

extracting a symbol from the received symbol sequence;

searching a lexicon with the extracted symbol and retrieving an entity-type corresponding to the symbol, the entity-type selected from the group consisting of a picture-like entity, a relation, and a picture-like entity accompanied by a relation;

grouping the symbols according to the retrieved entity-types;

repeating the extracting, searching, retrieving, and grouping steps for each symbol in the received symbol sequence; and

outputting said symbol grouping.

18. The method of claim 17, wherein the retrieving step comprises retrieving a cognitive category valuation for each symbol that designates a picture-like entity;

and wherein the method includes the step of inserting a relation-designating symbol between consecutive symbols in the received symbol sequence which designate picture-like entities.

19. The method of claim 17, wherein said grouping step includes building a symbol group consisting of picture-designating symbols and relation-designating symbols.

20. The method of claim 17, wherein said searching step further comprises retrieving symbol-grouping properties corresponding to the symbol;

and wherein said grouping step comprises grouping the inputted picture-designating and relation-designating symbols according to the retrieved entity-types and symbol-grouping properties.