System and method of using acoustic models for automatic speech recognition which distinguish pre- and post-vocalic consonants

A non-transitory computer-readable medium storing instructions for controlling a computing device to train a language model, the instructions comprising:

• receiving a speech signal;
• defining at least one syllable boundary position for a word in the speech signal;
• based on the at least one syllable boundary position, generating for each consonant in a syllable of the word one of a pre-vocalic position label and a post-vocalic position label;
• expanding a portion the consonant phoneme inventory associated with the word to include the one of the pre-vocalic position label and the post-vocalic position label to yield an expanded consonant phoneme inventory;
• reformulating a lexicon to reflect the expanded consonant phoneme inventory and yield a reformulated lexicon;
• and
• training language model for an automated speech recognition system based on the reformulated lexicon.

A system for improving output from a speech recognizer by training models, the system comprising;

• a processor;
• a first module configured to cause the processor to receive a speech signal;
• a second module configured to cause the processor to define at least one syllable boundary position for a word in the speech signal;
• a third module to cause the processor configured to generate for each consonant in a syllable of the word a pre-vocalic position label and a post-vocalic position label based on the at least one syllable boundary position;
• a fourth module configured to cause the processor to expand a portion of a consonant phoneme inventory associated with the word to include the one of the pre-vocalic position label and the post-vocalic position label to yield an expanded consonant phoneme inventory;
• a fifth module configured to cause the processor to reformulate a lexicon to reflect an expanded consonant phoneme inventory and yield a reformulated lexicon;
• and
• a sixth module configured cause the processor to train a language model for an automated speech recognition system based on the reformulated lexicon.