Topic Identification of News Speech based on Keyword Spotting

Bibliographic Information

Other Title
  • キーワードスポッティングに基づくニュース音声の話題同定

Search this article

Description

For efficient access to desired information from a large quantity of speech media, it is useful to index it with its topics. In this report, we study topic identification with keyword spotting and keyword selection, and apply them to news speech. For large vocabulary word spotting, we prepare two dictionaries. One is for keywords that contribute to the topic identification, and the other is for basic words that are used to approximate the input speech. For keyword selection, we adopt three criteria : mutual information with topics, similarity of the phone sequence, and its length. With the obtained dictionary sets, we have performed topic identification experiments both on news texts and speech.

Journal

  • IPSJ SIG Notes

    IPSJ SIG Notes 6 15-20, 1995-05-25

    Information Processing Society of Japan (IPSJ)

Citations (19)*help

See more

References(8)*help

See more

Details 詳細情報について

  • CRID
    1570291227261350400
  • NII Article ID
    110002916957
  • NII Book ID
    AN10442647
  • ISSN
    09196072
  • Text Lang
    ja
  • Data Source
    • CiNii Articles

Report a problem

Back to top