Speech processing apparatus, medium, and method recognizing and responding to speech using entity information

• US 8,185,397 B2
• Filed: 03/17/2006
• Issued: 05/22/2012
• Est. Priority Date: 03/29/2005
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Lee, Jaewon, Kang, Inho, Rim, Haechang, Kim, Jeongsu
• Primary Examiner(s): RIDER, JUSTIN W
• Application Number:
  US11/377,207
• Publication Number:
  US 20060224383A1
• Time in Patent Office: 2,258 Days
• Field of Search: 704/243, 704/270, 704/270.1
• US Class Current: 704/270
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning

Abstract

A speech processing apparatus, medium, and method recognizing speech and responding to the speech. The speech processing apparatus may includes an entity extracting unit which extracts entity information and an upper entity corresponding to the entity information from input speech, a focus determination unit which determines a focus using the extracted entity information requiring a response, a mapping unit which maps lower entity corresponding to the focus with the extracted entity information, and a recognition unit which recognizes a result of arranging the extracted entity information according to semantic association among the lower entities as the input speech. Thus, the speech processing apparatus can accurately recognize grammatically correct speech as well as grammatically incorrect speech and then respond to the speech.

Claims

1. A speech processing apparatus, comprising:

an entity extracting unit to extract entity information by extracting input speech entity information from input speech and extracting a respective upper entity from a respective same domain as the input speech entity information;

a focus determination unit to determine a focus of the input speech using extracted input speech entity information requiring a response from among a plurality of the extracted input speech entity information and preset entity information requiring a response, wherein the preset entity information requiring a response is pre-stored separately from other entity information;

a mapping unit to map lower domain entities, of the domain, corresponding to the focus with the extracted input speech entity information; and

a recognition unit to recognize a result of arranging the extracted input speech entity information according to semantic association among the mapped lower domain entities, of the domain, as the input speech,

wherein the upper entity is a hierarchically higher entity, within the domain, than the lower domain entities, the semantic association among the domain entities is preset, and entity portions within the extracted input speech entity information are determined to be semantically associated with one another according to the preset semantic association, and

wherein the recognition unit performs recognition of the input speech within pre-defined categories, respectively as the lower domain entities, with the lower domain entities being distinguished from the entity portions within the extracted input speech entity information and the recognized input speech.

2. The speech processing apparatus of claim 1, wherein the extracted input speech entity information comprises at least one individual word entity and/or at least one phrase entity.

3. The speech processing apparatus of claim 1, further comprising a response unit to respond to a recognized request according to the recognized result of the recognition unit.

4. The speech processing apparatus of claim 1, further comprising a verification unit to compare a degree of relation of the mapped lower domain entities with a preset reference level and to output a verification signal as a result of the comparison,

wherein the recognition unit recognizes the input speech in response to the verification signal.

5. The speech processing apparatus of claim 4, wherein the verification unit examines whether a combination of -selected plural lower domain entities, of the lower domain entities, are included in one of a preset plurality of combinations and outputs the verification signal as a result of the examination.

6. The speech processing apparatus of claim 1, wherein the entity extracting unit reads out the input speech entity information and the upper entity from a stored collection of the domain comprising preset entity information and a plurality of upper entities.

7. The speech processing apparatus of claim 1, wherein the entity information requiring the response includes entity information denoting a query or entity information denoting a command among the extracted input speech entity information.

8. The speech processing apparatus of claim 1, wherein the mapping unit reads the lower domain entities to be mapped from a stored collection of the domain comprising a preset plurality of lower domain entities.

9. A method of speech processing, the method comprising:

extracting entity information by extracting input speech entity information from input speech and extracting a respective upper entity from a respective same domain as the input speech entity information;

determining a focus of the input speech using extracted input speech entity information that requires a response from among a plurality of the extracted input speech entity information and preset entity information requiring a response, wherein the preset entity information requiring a response is pre-stored separately from other entity information;

mapping lower domain entities, of the domain, corresponding to the focus with the extracted input speech entity information; and

recognizing the extracted input speech entity information according to a semantic association among the mapped lower domain entities, in the domain, as the input speech,

wherein the upper entity is a hierarchically higher entity, within the domain, than the lower domain entities, the semantic association among the domain entities is preset, and entity portions within the extracted input speech entity information are determined to be semantically associated with one another according to the preset semantic association, and

wherein the recognizing of the extracted input speech entity information includes performing recognition of the input speech within pre-defined categories, respectively as the lower domain entities, with the lower domain entities being distinguished from the entity portions within the extracted input speech entity information and the recognized input speech.

10. The method of claim 9, wherein the extracted input speech entity information comprises at least one individual word entity and/or at least one phrase entity.

11. The method of claim 9, further comprising responding to a recognized request according to a result of the recognizing of the extracted input speech entity information.

12. The method of claim 9, further comprising comparing a degree of relation among the mapped lower domain entities with a preset reference level,

wherein, in the recognizing of the extracted input speech entity information, the input speech is recognized if the degree of relation is greater than the reference level.

13. The method of claim 12, wherein the comparing of the degree of relation further comprises determining whether a combination of selected plural lower domain entities form one of a plurality of preset combinations, and

in the recognizing of the extracted entity information, the input speech is recognized if the combination of the selected plurality of lower domain entities, of the lower domain entities, form one of the plurality of preset combinations.

14. The method of claim 9, wherein, in the extracting of the input speech entity information and the upper entity, the input speech entity information and the upper entity are read from a stored collection of the domain comprising preset entity information and upper entities.

15. The method of claim 9, wherein, in the mapping of the lower domain entities, the lower domain entities to be mapped are read from a stored collection of the domain comprising a plurality of preset lower domain entities.

16. At least one computer readable recording medium storing instructions that control at least one processor to perform a method of speech processing, the method comprising:

extracting entity information by extracting input speech entity information from input speech and extracting a respective upper entity from a respective same domain as the input speech entity information;

determining a focus of the input speech using extracted input speech entity information that requires a response from among a plurality of the extracted input speech entity information and preset entity information requiring a response, wherein the preset entity information requiring a response is pre-stored separately from other entity information;

mapping lower domain entities, of the domain, corresponding to the focus with the extracted input speech entity information; and

recognizing the extracted input speech entity information according to a semantic association among the mapped lower domain entities, in the domain, as the input speech,

wherein the upper entity is a hierarchically higher entity, within the domain, than the lower domain entities, the semantic association among the domain entities is preset, and entity portions within the extracted input speech entity information are determined to be semantically associated with one another according to the preset semantic association, and

wherein the recognizing of the extracted input speech entity information includes performing recognition of the input speech within pre-defined categories, respectively as the lower domain entities, with the lower domain entities being distinguished from the entity portions within the extracted input speech entity information and the recognized input speech.