Multi-class composite N-gram language model

Hirofumi Yamamoto, Shuntaro Isogai, Yoshinori Sagisaka

Research output: Contribution to journalArticle

35 Citations (Scopus)

Abstract

A new language model is proposed to cope with the scarcity of training data. The proposed multi-class N-gram achieves an accurate word prediction capability and high reliability with a small number of model parameters by clustering words multi-dimensionally into classes, where the left and right context are independently treated. Each multiple class is assigned by a grouping process based on the left and right neighboring characteristics. Furthermore, by introducing frequent word successions to partially include higher order statistics, multi-class N-grams are extended to more efficient multi-class composite N-grams. In comparison to conventional word tri-grams, the multi-class composite N-grams achieved 9.5% lower perplexity and a 16% lower word error rate in a speech recognition experiment with a 40% smaller parameter size.

Original languageEnglish
Pages (from-to)369-379
Number of pages11
JournalSpeech Communication
Volume41
Issue number2-3
DOIs
Publication statusPublished - 2003 Oct 1
Externally publishedYes

Keywords

  • Class N-gram
  • N-gram language model
  • Variable length N-gram
  • Word clustering

ASJC Scopus subject areas

  • Software
  • Modelling and Simulation
  • Communication
  • Language and Linguistics
  • Linguistics and Language
  • Computer Vision and Pattern Recognition
  • Computer Science Applications

Fingerprint Dive into the research topics of 'Multi-class composite N-gram language model'. Together they form a unique fingerprint.

  • Cite this