/ Language Modeling / Tutorial Book / Tutorials / Software / Home
Overview:
black fading line

Previous sections of the tutorial have explained how to extract features from speech and how to use them to build acoustic models. The speech recognizer references these models to determine phonemes that comprise words. The language model provides additional knowledge for the recognizer by specifying the order in which those words are likely to occur. Early attempts at language modeling used isolated word recognition, in which the speaker was required to pause after each word spoken. Modern recognizers can decode continuous speech, consisting of sequences of words that are not necessarily separated by a pause. Two popular language models used to build these recognizers include Network and N-gram. The ISIP software supports development of either of these language models. Continue to Section 6.1.1 for further theoretical overview.

Contents:


6.1 Overview
  6.1.1 Introduction
  6.1.2 Bayes' Rule
  6.1.3 Grammar Definition
  6.1.4 Chomsky's Hierarchy
  6.1.5 Grammars for Speech Recognition
 
6.2 Language Modeling Tools
  6.2.1 Network Builder
  6.2.2 Basic User Interface
  6.2.3 Building a Simple Network
  6.2.4 Generating an Acoustic Model
  6.2.5 Language Model Tester
  6.2.6 SRI Toolkit
6.3 N-Gram Modeling
  6.3.1 Overview
  6.3.2 Training an N-gram Language Model
  6.3.3 N-Gram Language Model Structure
  6.3.4 N-Gram Decoding
  6.3.5 Scoring
   
Table of Contents   Section Contents   Previous Page Up Next Page
      Glossary / Help / Support / Site Map / Contact Us / ISIP Home