Comparison of Word and Subword Indexing Techniques for Mandarin Chinese Spoken Document Retrieval

In this paper, we investigate the use of words and subwords (including both characters and syllables) in audio indexing for Mandarin Chinese spoken document retrieval. Two retrieval approaches, including the well-known vector space model approach and the newly proposed HMM/Ngram-based approach, are used in the present work. We focus on the use of an entire Chinese textual story (from a newspaper) as a query to retrieve Mandarin Chinese spoken documents (from news broadcasts). Experiments are based on the Topic Detection and Tracking Corpora.

• Type: Conference
• Year: 2001
• Where: PCM
• Authors: Hsin-Min Wang, Berlin Chen
• Keywords: Chinese Spoken Documents, Mandarin Chinese, Multimedia, PCM 2001, Spoken Document Retrieval