Speech retrieval method, speech retrieval apparatus, and program for speech retrieval apparatus

• US 9,373,328 B2
• Filed: 06/22/2015
• Issued: 06/21/2016
• Est. Priority Date: 04/21/2014
• Status: Active Grant

• Current Assignee: SINOEAST CONCEPT LIMITED (Tencent Holdings Limited)
• Original Assignee: International Business Machines Corporation
• Inventors: Kurata, Gakuto, Nagano, Tohru, Nishimura, Masafumi
• Primary Examiner(s): ALBERTALLI, BRIAN LOUIS
• Application Number:
  US14/745,912
• Publication Number:
  US 20150310860A1
• Time in Patent Office: 365 Days
• US Class Current: 1/1
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/04 Segmentation; Word boundary detection
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units
• CPC Class Codes: G10L 2015/027 Syllables being the recognition units
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 25/51 for comparison or discrimination

Abstract

A method for speech retrieval includes acquiring a keyword designated by a character string, and a phoneme string or a syllable string, detecting one or more coinciding segments by comparing a character string that is a recognition result of word speech recognition with words as recognition units performed for speech data to be retrieved and the character string of the keyword, calculating an evaluation value of each of the one or more segments by using the phoneme string or the syllable string of the keyword to evaluate a phoneme string or a syllable string that is recognized in each of the detected one or more segments and that is a recognition result of phoneme speech recognition with phonemes or syllables as recognition units performed for the speech data, and outputting a segment in which the calculated evaluation value exceeds a predetermined threshold.

Claims

1. A method for speech retrieval, comprising:

acquiring a keyword designated by a character string, and a phoneme string or a syllable string;

detecting one or more coinciding segments by comparing a character string that is a recognition result of word speech recognition with words as recognition units performed for speech data to be retrieved and the character string of the keyword;

calculating an evaluation value of each of the one or more coinciding segments by using the phoneme string or the syllable string of the keyword to evaluate a phoneme string or a syllable string that is recognized in each of the detected one or more coinciding segments and that is a recognition result of phoneme speech recognition with phonemes or syllables as recognition units performed for the speech data, wherein the phoneme string or the syllable string associated with each of the coinciding segments is a phoneme string or a syllable string associated with a segment in which a start and an end of the segment is expanded by a predetermined time; and

outputting a segment in which the calculated evaluation value exceeds a predetermined threshold.

2. The method according to claim 1, wherein calculating comprises:

comparing a phoneme string or a syllable string that is an N-best recognition result of phoneme speech recognition with phonemes or syllables as recognition units performed for speech data associated with each of the detected one or more coinciding segments and the phoneme string of the keyword to set a rank of the coinciding N-best recognition result as the evaluation value.

3. The method according to claim 1, wherein calculating comprises:

setting, as the evaluation value, an edit distance between a phoneme string or a syllable string that is a 1-best recognition result of phoneme speech recognition with phonemes or syllables as recognition units performed for speech data associated with each of the detected one or more coinciding segments and the phoneme string or the syllable string of the keyword.

4. The method according to claim 3, wherein the edit distance is a distance matched by matching based on dynamic programming.

5. The method according to claim 1, further comprising performing word speech recognition of the speech data to be retrieved, with words as recognition units.

6. The method according to claim 1, further comprising performing phoneme speech recognition of the speech data associated with each of the detected one or more coinciding segments, with phonemes or syllables as recognition units.

7. The method according to claim 1, further comprising performing phoneme speech recognition of the speech data to be retrieved, with phonemes or syllables as recognition units.

8. The method of claim 1, wherein the calculating the evaluation value of each of the one or more coinciding segments further includes using the character string of the keyword to evaluate the character string in each of the detected one or more coinciding segments.

9. The method of claim 1, further comprising adjusting the predetermined threshold to alter at least one of a precision value and a recall value of the output segment, the precision value being positively correlated with the predetermined threshold and the recall value being negatively correlated with the predetermined threshold.

10. The method of claim 9, wherein the precision value is a ratio of retrieval results satisfying a retrieval request to all documents satisfying the retrieval request.

11. The method of claim 9, wherein the recall value is a ratio of retrieval results satisfying a retrieval request to all retrieval results.