Abstract:
This paper approaches and implements a retrieval engine for speech data based on relevance feedback, imposing Sphinx system to transfer audio file to text file then indexed by Lucene system. To improve the quality of the query answering, the paper employs an overall relevance feedback, including local feedback and global feedback. It improves the system’s accuracy and dynamic characteristic. Experimental results prove the engine meets the users’ demands and it’s of practice value.
Key words:
relevance feedback,
speech retrieval engine,
Sphinx,
Lucene
摘要: 提出并实现了一种基于相关反馈的语音检索引擎,该引擎基于Sphinx语音识别工具将语音转化为文本,再采用Lucene对文本进行索引。为了提高语音检索的质量,系统引入了相关反馈机制,不仅通过局部相关反馈修正用户的查询,还通过全局类相关反馈机制挖掘Sphinx的识别错误模式,扩展了用户的查询,大大增强了该索引系统的准确性和实时动态性。实验结果证明该系统能符合检索者的需求,具有实用价值。
关键词:
相关反馈,
语音检索引擎,
Sphinx,
Lucene
CLC Number:
YE Liang; WANG Zhi-bin; SHAO Qian-ming. Speech Retrieval Engine Based on Relevance Feedback[J]. Computer Engineering, 2007, 33(17): 228-230.
叶 靓;王智斌;邵谦明. 基于相关反馈的语音检索引擎[J]. 计算机工程, 2007, 33(17): 228-230.