In this work we concentrate on generating the compound words with high order n-gram information for speech recognition. It is reported that the long phrases in vocabulary are more probable to appear during the decoding task. In most existed methods, only bi-gram information is under the consideration within the constraint of the computational resources and the much longer compound words are generated in an iterative way. However, many long phrases can not be iteratively built and the bi-gram information can only provide very limited help when 4-gram Language model is used during decoding. Here we present a new form of generation criterion and separate it into prediction part and history part. This largely saves the computational cost and can be extended to any higher order cases. In our experiment on mandarin Open Voice Search (OVS) work we make 0:62% percents absolute improvement and outperform the traditional mutual information based methods.
By: Jie Zhou, Qin Shi, Yong Qin
Published in: RC25063 in 2010
LIMITED DISTRIBUTION NOTICE:
This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.
Questions about this service can be mailed to reports@us.ibm.com .