基于音素混淆模型的集外词查询项扩展方法

为了提高语音查询项检索系统中集外词检索的性能,在加权有限状态转换器（weightedfinite-state transducer,WFST）框架下提出了一种基于音素混淆模型的集外词查询项扩展技术,将查询项扩展成多发音序列来解决集外词问题.首先由G2P（grapheme-to-phoneme）模型生成查询项的发音序列,然后利用音素混淆模型将发音序列扩展成N-best发音,以补偿识别错误造成Lattice建立的索引与查询项发音序列之间音素表示差异带来的影响,从而有效降低漏警率.实验结果表明,加入音素混淆模型之后,系统集外词检索性能有明显提升.

Query Expansion Method for Out-of-Vocabulary Based on Phonetic Confusion Model

To improve the performance of spoken term detection systems, a query expansion method for out-of-vocabulary （OOV） based on phonetic confusion model is presented in the weighted finite- state transducer framework （WFST）. The problem of OOV is solved by expanding the queries to multiple pronunciation sequences. First, a pronunciation sequence is generated by grapheme-to-pho- neme model; then, the pronunciation sequence is expanded to N-best sequences by phonetic confusion model to compensate for potential differences caused by recognition errors in deriving index and query representations, thus reducing the missing alarm rate effectively. The experimental results show that the OOV retrieval performance of the system is improved significantly by the expansion based on phonetic confusion model.