A New Framework to Deal with OOV Words in SLT System

Automatic spoken language translation (SLT) is considered as one of the most challenging tasks in modern computer science and technology. It is always a hard nut to deal with the problem of Out-Of-Vocabulary (OOV) words in SLT. The existing traditional SLT framework often doesn’t take effect for OOV words translation because of the data sparseness. In this paper based on the analysis of common OOV expressions appeared in SLT, we propose a new framework for bidirectional Chinese-English SLT in which a series of approaches to translating OOV expressions are presented. The experimental results have shown that our framework and approaches are effective and can greatly improve the translation performance.