Azure Cognitive Speech Service

You can make customised language model using Customer Speech Service and sound model which is well connected with your user and your application. You can upload specific speech/text data for customised models as per your need that can combine with existing art of speech models by Microsoft.

Working

Several components work together in Speech Recognition System. Two most component are sound model and language model.

Acoustic or sound models

Short fragments of audio are labelled into one of phonemes or sound units to a given language. Example, word speech is made up of four phonemes “ s p iy ch”.

Language model

Sequence of word sound similar, these model help to decide the word to system. For example, the word ‘know’ and ‘no’ sounds similar. Both the model statistics are learned from training data. So, they perform best when encounter the speech which is similar to training data.

UtterlyCloud

• admin@utterlycloud.com
• India