Position-dependent phonetic models for reliable pronunciation identification

• US 8,135,590 B2
• Filed: 01/11/2007
• Issued: 03/13/2012
• Est. Priority Date: 01/11/2007
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Liu, Peng, Shi, Yu, Soong, Frank Kao-ping
• Primary Examiner(s): Vo, Huyen X.
• Application Number:
  US11/652,451
• Publication Number:
  US 20080172224A1
• Time in Patent Office: 1,888 Days
• Field of Search: 704/254, 704/255, 704/220, 704/1, 704 3- 4, 704 6- 7, 704 9- 10, 704/251, 704/231, 704/243, 704/244, 704/270, 704/242, 704/256.4
• US Class Current: 704/255
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams

Abstract

A representation of a speech signal is received and is decoded to identify a sequence of position-dependent phonetic tokens wherein each token comprises a phone and a position indicator that indicates the position of the phone within a syllable.

Claims

1. A method comprising:

receiving a representation of a speech signal;

a processor decoding the representation of the speech signal into a sequence of words using a word language model and an acoustic model;

a processor converting each word in the sequence of words into a sequence of position-dependent phonetic tokens using a phonetic token lexicon, which provides a position-dependent phonetic token description of each word in a lexicon, wherein each position-dependent phonetic token comprises a phone and a position indicator that indicates the position of the phone within a syllable; and

a processor determining probabilities for sub-sequences in the sequences of position-dependent phonetic tokens by applying sub-sequences of position-dependent phonetic tokens converted from the sequence of words to a position-dependent phonetic language model that describes probabilities of sequences of position-dependent phonetic tokens comprising a conditional probability of a position-dependent phonetic token given at least two preceding position-dependent phonetic tokens and probabilities of individual position-dependent phonetic tokens.

2. The method of claim 1 further comprising utilizing the at least one probability as a confidence measure.

3. The method of claim 1 wherein at least one position indicator indicates one of a group of syllable positions consisting of onset consonant, and coda consonant.

4. A hardware computer storage medium encoded with a computer program, causing the computer to execute steps comprising:

decoding a representation of a speech signal using a position-dependent phonetic language model that provides probabilities of sequences of position-dependent phonetic tokens wherein each position-dependent phonetic token comprises a phone and a position identifier that identifies a position within a syllable, wherein decoding produces at least one sequence of position-dependent phonetic tokens;

receiving a symbol represented by a portion of the speech signal, wherein the symbol is not part of the language of a lexicon; and

annotating a lexicon by storing a sequence of position-dependent phonetic tokens as a pronunciation for the symbol that is not part of the language of the lexicon, wherein annotating the lexicon further comprises identifying syllable boundaries from the position-dependent phonetic tokens and placing syllable boundaries in the lexicon.

5. The hardware computer storage medium of claim 4 wherein a position identifier is one of a group of position identifiers consisting of onset consonant and coda consonant.