Device for extracting keywords in a conversation

• US 8,370,145 B2
• Filed: 03/14/2008
• Issued: 02/05/2013
• Est. Priority Date: 03/29/2007
• Status: Expired due to Fees

• Current Assignee: Sovereign Peak Ventures, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Panasonic Corporation (Panasonic Holdings Corporation)
• Inventors: Endo, Mitsuru, Yamada, Maki, Morii, Keiko, Konuma, Tomohiro, Nomura, Kazuya
• Primary Examiner(s): Smits, Talivaldis Ivars
• Assistant Examiner(s): ROBERTS, SHAUN A
• Application Number:
  US12/302,633
• Publication Number:
  US 20090150155A1
• Time in Patent Office: 1,789 Days
• Field of Search: 704/255, 704/231, 704/275, 704/235, 704/251, 379/88.24
• US Class Current: 704/255
• CPC Class Codes: G10L 15/04 Segmentation; Word boundary detection
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

The present invention aims at extracting a keyword of conversation without preparations by advanced anticipation of keywords of conversation. A keyword extracting device of the present invention includes an audio input section 101 by way of which a speech sound made by a speaker is input; a speech segment determination section 102 that determines a speech segment for each speaker in connection with the input speech sound; a speech recognition section 103 that recognizes a speech sound of the determined speech segment for each speaker; an interrupt detection section 104 that detects a feature of a speech response suggesting presence of a keyword on the basis of a response of another speaker to speech sounds of respective speakers; namely, an interrupt where a preceding speech and a subsequent speech overlap; a keyword extraction section 105 that extracts the keyword from the speech in the speech segment specified on the basis of an interrupt; a keyword search section 106 that performs keyword search by means of the keyword; and a display section 107 that displays a result of keyword search.

Claims

1. A keyword extracting device, comprising:

an audio input section that inputs speech sound of a first speaker and a second speaker, wherein at least a start of the speech sound of the first speaker precedes a start of the speech sound of the second speaker, and the speech sound of the first and second speakers includes speech sound of a preceding speech and speech sound of a subsequent speech;

a speech segment determination section that determines respective first and second speech segments respectively corresponding to the first and second speakers in connection with the speech sound so as to separately identify speech sound from the first and second speakers;

a speech recognition section that recognizes respective speech sounds of the respective first and second speech segments determined for the first and second speakers;

a speech response feature extraction section that detects a speech response feature only from the second speech segment corresponding to the second speaker, the speech response feature indicates a presence of a keyword in the first speech segment corresponding to the first speaker based on the second speech segment as a response to the first speech segment of the first speaker; and

a keyword extraction section that extracts the keyword from the first speech segment specified by the speech response feature detected by the speech response feature extraction section, wherein the keyword extraction section extracts, as the keyword, a constituent element at the end of the first speech segment corresponding to the preceding speech prior to the speech response.

2. The keyword extracting device according to claim 1,

wherein the speech response feature extraction section includes an interrupt detection section that detects, on the basis of the first and second speech segments, an interrupt where the preceding speech and the subsequent speech overlap each other, when the second speech segment commences in a middle of the first speech segment;

and

wherein the keyword extraction section extracts the keyword from the first speech segment corresponding to the preceding speech when the interrupt is detected.

3. The keyword extracting device according to claim 1,

wherein the speech response feature extraction section includes:

a pitch determination section that determines respective pitches of the speech sound of the preceding speech and the speech sound of the subsequent speech; and

a pattern determination section that determines, from the determined pitches, whether a pitch pattern including a descending pitch at an end of the preceding speech and an ascending pitch at a beginning of the subsequent speech is indicated; and

wherein the keyword extraction section extracts the keyword from the first speech segment corresponding to the preceding speech when the pitch pattern is indicated.

4. The keyword extracting device according to claim 1,

wherein the speech response feature extraction section detects a functional phrase of a predetermined type the second speech segment corresponding to the subsequent speech;

and

wherein the keyword extraction section extracts the keyword from the first speech segment corresponding to the preceding speech which precedes the second speech segment, corresponding to the subsequent speech, that includes the detected functional phrase.

5. The keyword extracting device according to claim 1, wherein the speech response feature extraction section detects an exciting reaction of a person, other than the first and second speakers,located in a vicinity of the first and second speech segments respectively determined for the first and second speakers;

and

wherein the keyword extraction section extracts the keyword from one of the first or second speech segments detected as corresponding to the exciting reaction.

6. The keyword extracting device according to claim 1, wherein the speech response feature extraction section recognizes facial expression of another speaker responsive to speech sounds of the first and second speakers and detects a point of change in the recognized facial expression;

and

wherein, prior to the speech response, the keyword extraction section extracts, as a keyword, a constituent element at the end of the first speech segment corresponding to the detected point of change in facial expression.