Speech recognition accuracy via concept to keyword mapping

• US 7,925,506 B2
• Filed: 10/05/2004
• Issued: 04/12/2011
• Est. Priority Date: 10/05/2004
• Status: Expired due to Fees

• Current Assignee: iNAGO Corporation
• Original Assignee: iNAGO Corporation
• Inventors: Farmaner, Gary, Dicarlantonio, Ron, Leonard, Huw
• Primary Examiner(s): Wozniak, James S
• Assistant Examiner(s): SHAH, PARAS D
• Application Number:
  US10/957,579
• Publication Number:
  US 20060074671A1
• Time in Patent Office: 2,380 Days
• Field of Search: 704/257, 704/255, 704/251, 704/231, 704/246
• US Class Current: 704/257
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/193 Formal grammars, e.g. finite state automata, context free grammars or word networks

Abstract

The invention provides a system and method for improving speech recognition. A computer software system is provided for implementing the system and method. A user of the computer software system may speak to the system directly and the system may respond, in spoken language, with an appropriate response. Grammar rules may be generated automatically from sample utterances when implementing the system for a particular application. Dynamic grammar rules may also be generated during interaction between the user and the system. In addition to arranging searching order of grammar files based on a predetermined hierarchy, a dynamically generated searching order based on history of contexts of a single conversation may be provided for further improved speech recognition. Dialogue between the system and the user of the system may be recorded and extracted for use by a speech recognition engine to refine or create language models so that accuracy of speech recognition relevant to a particular knowledge area may be improved.

Claims

1. A speech recognition system for providing a textual output from an audible signal representative of spoken words, said system comprising:

a storage unit, said storage unit storing a plurality of sentence types on a storage medium, each of said sentence type including a pre-arranged ordered list of concepts, each concept in said pre-arranged ordered list of concepts being expandable to a plurality of words conceptually related to the each expanded concept;

a natural language processor operatively connected to said storage unit, said natural language processor being configured to parse a partially recognized sentence having an ordered list of recognized words and unrecognized sound groupings into a first ordered list of concepts and placeholders, said concepts in said first ordered lists corresponding to said recognized words and said placeholders corresponding to said unrecognized sound groupings, and said natural language processor being further configured to identify a sentence type from said plurality of sentence types utilizing said first ordered list of concepts and placeholders, said identified sentence type having an associated ordered list of concepts, the order of concepts in said associated ordered list of concepts being determined by the spoken words, each concept in said associated ordered list of concepts being expandable to a plurality of words conceptually related to the each expanded concept, said partially recognized sentence corresponding to the audible signal;

a grammar rule generator for expanding each of said concepts at a location corresponding to one of said unrecognized sound groupings into a respective plurality of related words;

a speech recognition engine for converting the audible signal to the textual output, said speech recognition engine being operatively connected to said respective plurality of related words for resolving the one of said unrecognized sound grouping.

2. The speech recognition system of claim 1, wherein the storage unit stores a collection of concepts and mappings from concepts to keywords, said storage unit being further operatively connected to the grammar rule generator.

3. The speech recognition system of claim 1, further comprising a language model module operatively connected to the speech recognition engine, said language model module being configured to assist the speech recognition engine to convert the audible signal and to receive the recognized textual output from the speech recognition engine to add to a corpus for creating and refining language models.

4. The system of claim 1, wherein said associated ordered list of concepts has a context designation assigned thereto.

5. The system of claim 4, wherein said system is operable to determine a conversation context of the spoken words, and said natural language processor is operable to select said associated ordered list of concepts if said context designation matches said conversation context.

6. A system for improving recognition accuracy of an audible signal representative of spoken words, the audible signal being converted to a textual output by a speech recognition engine, said system comprising:

a storage unit, said storage unit storing a plurality of pre-arranged ordered lists of keywords on a storage medium, each keyword in said each pre-arranged ordered list of keywords being expandable to a plurality of words conceptually related to the each expanded keyword;

a natural language processor operatively connected to said storage unit, said natural language processor being configured to parse a sentence in a textual format into an ordered list of keywords identified from said pre-arranged ordered lists of keywords, the order of keywords in said identified ordered list of keywords being determined by the sentence, each keyword in said identified ordered list of keywords being expandable to a plurality of words conceptually related to the each expanded keyword; and

a grammar rule generator for expanding each keyword of said identified ordered list into a respective plurality of related words to obtain a grammar rule from said ordered list of keywords;

wherein said speech recognition engine is operatively connected to said grammar rule for resolving unrecognized sound groupings in the audible signal into the corresponding spoken words in the textual output.

7. The system of claim 6, further comprising an editor for preparing concept to keywords mappings, wherein said expansion of each keyword into said respective plurality of related words corresponds to matching each said keyword to a concept and replacing said concept with keywords using a corresponding concept to keywords mapping.

8. The system of claim 6, wherein said grammar rule has a context designation assigned thereto.

9. The system of claim 8, wherein said system is operable to determine a conversation context of the spoken words, and said speech recognition engine is operable to select said grammar rule if said context designation matches said conversation context.