Speech recognition with feedback from natural language processing for adaptation of acoustic models

• US 6,879,956 B1
• Filed: 09/29/2000
• Issued: 04/12/2005
• Est. Priority Date: 09/30/1999
• Status: Expired due to Term

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Honda, Hitoshi, Omote, Masanori, Ogawa, Hiroaki, Pao, Hongchang
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Lerner, Martin
• Application Number:
  US09/676,644
• Time in Patent Office: 1,656 Days
• Field of Search: 704/231, 704/243, 704/244, 704/251, 704/255, 704/257, 704/1, 704/9
• US Class Current: 704/244
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/075 supervised, i.e. under machine guidance
• CPC Class Codes: G10L 2015/0638 Interactive procedures

Abstract

A speech recognition unit performs speech recognition on input speech based on models, such as HMMs, and supplies a speech recognition result to a dialog management unit. The dialog management unit forms a reply to the speech recognition result. In this case, the dialog management unit detects zones in which the speech recognition result is correct based on the speech recognition result and the associated reply, and feeds them back to the speech recognition unit. Then, the speech recognition unit performs on-line adaptation of the models based on the speech of the correct zones fed back from the dialog management unit and the associated speech recognition result.

Claims

1. A speech processing apparatus comprising:

speech recognition means for performing speech recognition on input speech to produce a speech recognition result using acoustic models; and

natural-language processing means for performing natural language processing on said speech recognition result, said natural-language processing means including

feedback means for feeding back information obtained as a result of the natural language processing performed on the speech recognition result to said speech recognition means, said feedback information including adaptation zones,

wherein said speech recognition means comprises

adaptation process means for processing the information fed back from said feedback means to adapt the acoustic models so that said speech recognition means produces the speech recognition result with higher precision than when said adaptation process means is not used.

2. A speech processing apparatus according to claims 1,

wherein said feedback means feeds back at least one of said adaptation zones which are to be used for the adaptation of the acoustic models and said adaptation zones which are not to be used for the adaptation of the acoustic models.

3. A speech processing apparatus according to claims 1,

wherein said feedback means feeds back the speech recognition result which appears to be correct.

4. A speech processing apparatus according to claim 1,

wherein said feedback means feeds back the reliability of the speech recognition result.

5. A speech processing apparatus according to claim 1,

wherein said feedback means feeds back a task of the speech recognition result.

6. A speech processing apparatus according to claim 1,

wherein said feedback means comprising feeding back at least one of

(a) said adaptation zones which are to be used for the adaptation of the acoustic models,

(b) said adaptation zones which are not to be used for the adaptation of the acoustic models,

(c) the speech recognition result which appears to be correct,

(d) the reliability of the speech recognition result, and

(e) a task of the speech recognition result.

7. A speech processing method comprising:

a speech recognition step of performing speech recognition on input speech to produce a speech recognition result using acoustic models; and

a natural-language processing step of performing natural language processing on said speech recognition result, said natural-language processing step including

a feedback step of feeding back information obtained as a result of the natural language processing performed on the speech recognition result to said speech recognition step, said feedback information including adaptation zones,

wherein said speech recognition step comprises

an adaptation process step of processing the information fed back from said feedback step to adapt the acoustic models so that said speech recognition step produces the speech recognition result with higher precision than when said adaptation process step is not performed.

8. A recording medium for recording a program which causes a computer to perform speech recognition processing, said program comprising:

a speech recognition step of performing speech recognition on input speech to produce a speech recognition result using acoustic models; and

a natural-language processing step of performing natural language processing on said speech recognition result, said natural-language processing step including

a feedback step of feeding back information obtained as a result of the natural language processing performed on the speech recognition result to said speech recognition step, said feedback information including adaptation zones,

wherein said speech recognition step comprises

an adaptation process step of processing the information fed back from said feedback step to adapt the acoustic models so that said speech recognition step produces the speech recognition result with higher precision than when said adaptation process step is not performed.