Topic Identification of News Speech using Word Cooccurrence Statistics

8 Citations 5 References

Bibliographic Information

Other Title
  • 単語の共起情報を用いたニュース朗読音声の話題同定機構

Search this article

Description

We are developing a retrieval system of news speech by indexing the audio data with their topics. Each keyword has a statistic of topic contribution extracted from the text database of the newspaper. By summing up these values of the recognized keywords, we identify the topics of the news speech sentences. For more reliable topic identification we propose the use of statistics of word cooccurrenses as well as those of individual words. We evaluate the effectiveness of this mechanism by the simulation.

Journal

  • IEICE technical report. Speech

    IEICE technical report. Speech 96 (449), 71-78, 1997-01-17

    The Institute of Electronics, Information and Communication Engineers

Citations (8)*help

See more

References(5)*help

See more

Keywords

Details 詳細情報について

  • CRID
    1573105977177347072
  • NII Article ID
    110003296324
  • NII Book ID
    AN10013221
  • Text Lang
    ja
  • Data Source
    • CiNii Articles

Report a problem

Back to top