Confusable word detection in speech recognition

• US 5,737,723 A
• Filed: 08/29/1994
• Issued: 04/07/1998
• Est. Priority Date: 08/29/1994
• Status: Expired due to Term

• Current Assignee: Alcatel-Lucent USA, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Roe, David Bjorn, Riley, Michael Dennis
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Chawan, Vijay B.
• Application Number:
  US08/297,283
• Time in Patent Office: 1,317 Days
• Field of Search: 395/2.52, 395/2.53, 395/2.54, 395/2.56, 395/2.61, 395/2.49
• US Class Current: 704/243
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams

Abstract

A speech recognition system may be trained with data that is independent from previous acoustics. This method of training is quicker and more cost effective than previous training methods. In training the system, after a vocabulary word is input into the system, a first set of phonemes representative of the vocabulary word is determined. Next, the first set of phonemes is compared with a second set of phonemes representative of a second vocabulary word. The first vocabulary word and the second vocabulary word are different. The comparison generates a confusability index. The confusability index for the second word is a measure of the likelihood that the second word will be mistaken as another vocabulary word, e.g., the first word, already in the system. This process may be repeated for each newly desired vocabulary word.

Claims

1. A method comprising the steps of:

(a) inputting a first vocabulary word in a form independent of previous acoustics into a speech recognition unit training system;

(b) determining a first set of phonemes representative of the first vocabulary word; and

(c) comparing the first set of phonemes with a second set of phonemes representative of a second vocabulary word to determine a confusability index, the first vocabulary word and the second vocabulary word being different.

2. The method of claim 1 wherein the first set of phonemes is comprised of a first set of phonemic transcriptions and the second set of phonemes is comprised of a second set of phonemic transcriptions.

3. The method of claim 2 wherein the step of inputting the first vocabulary word in a form independent of previous acoustics comprises typing in a first vocabulary word.

4. The method of claim 2 wherein the step of inputting the first vocabulary word in a form independent of previous acoustics comprises reading in the vocabulary word from a text file.

5. The method of claim 1 further comprising the steps of:

(a) determining a relationship between the confusability index and a threshold value;

(b) if the relationship may by characterized a particular way, storing the first set of phonemes; and

(c) determining if other vocabulary words need to be input.

6. A method of training a speech recognition system comprising:

(a) receiving a signal representing a first set of one or more characters;

(b) converting the first set to an acoustic representation of the first set;

(c) comparing the acoustic representation of the first set with a second acoustic representation to determine a confusability index, the second acoustic representation being of a second set of one or more characters, the first set of characters and the second set of characters being different.