Networked Information Retrieval Using Unconstrained Mandarin Speech Queries

Abstract

This paper presents new research on Chinese networked information retrieval with the capabilities of processing unconstrained speech queries and providing efficient natural language searching. The present system is designed to be able to resolve the conventional difficulties of both the input of Chinese characters and the retrieval of large Chinese texts in terms of network information access. In dealing with these difficulties, a signature-based natural language search method has been developed. Meanwhile, for handling of unconstrained speech queries , the present system is integrated with continuous Mandarin speech recognition technology and other robust techniques. With the distinguished features of the present system, the preliminary experimental results show that it achieves good performance in many ways, especially in terms of the accuracy of retrieval, the flexibility of query formation and the possibility of developing a spoken dialogue interface.