An Impovement for Adaptiong the Alignment Template Based Statistical Machine Translation Model to Pervasive Computing Environments

Machine translation (MT) has a variety of applications in pervasive computing environments. Thin clients such as personal digital assistants (PDA) are always deployed with small memory. In this paper, a novel method to improve the performance of alignment template based translation model (ATTM), which is the state-of-art MT models, was proposed. The key element, alignment template of ATTM, was extended to be context sensitive alignment templates (CSAT) based on the context information. After applying CSAT, the ATTM would use less time and memory during the decoding stage, and thus might be more suitable for small memory environments. In the end of this paper, an experiment was presented and the results showed that this proposed method could improve the performance greatly