Using word confidence score, insertion and substitution thresholds for selected words in speech recognition

• US 9,478,218 B2
• Filed: 10/24/2008
• Issued: 10/25/2016
• Est. Priority Date: 10/24/2008
• Status: Active Grant

• Current Assignee: Adacel, Inc. (Adacel Technologies Ltd.)
• Original Assignee: Adacel, Inc. (Adacel Technologies Ltd.)
• Inventors: Shu, Chang-Qing
• Primary Examiner(s): Godbold, Douglas
• Application Number:
  US12/258,093
• Publication Number:
  US 20100106505A1
• Time in Patent Office: 2,923 Days
• Field of Search: 704231-245
• US Class Current: 1/1
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 25/51 for comparison or discrimination

Abstract

A method and system for improving the accuracy of a speech recognition system using word confidence score (WCS) processing is introduced. Parameters in a decoder are selected to minimize a weighted total error rate, such that deletion errors are weighted more heavily than substitution and insertion errors. The occurrence distribution in WCS is different depending on whether the word was correctly identified and based on the type of error. This is used to determine thresholds in WCS for insertion and substitution errors. By processing the hypothetical word (HYP) (output of the decoder), a mHYP (modified HYP) is determined. In some circumstances, depending on the WCS'"'"'s value in relation to insertion and substitution threshold values, mHYP is set equal to: null, a substituted HYP, or HYP.

Claims

1. A method for recognizing speech in acoustic data, comprising:

performing a tuning phase, the tuning phase further comprising;

generating a series of hypothetical words (HYP) from a tuning audio data set in a decoder; and

setting values of tunable parameters in the decoder to minimize a weighted total error rate (Wt Etotal);

wherein the weighted total error is calculated according to an algorithm;

Wt Etotal=(λ

sub*num_error_sub_word+λ

ins*num_error_ins_word+λ

del*num_error_del_word)/total_num_RefWord,

where λ

sub, λ

ins, and λ

del are weighting factors, at least two of which are different,

num_error_sub_word is a number of substitution word errors,

num_error_ins_word is a number of insertion word errors,

num_error_del_word is a number of deletion word errors, and

total_num_RefWord is the number of words in the transcription.

2. The method of claim 1 wherein λ

del>

λ

sub>

λ

ins.