A survey of statistical machine translation using paraphrasing technology
Yidong Chen · Caai Transactions on Intelligent Systems · 2013
In this paper,the research team discussed possible new prospective research directions of paraphrasing technology in statistical machine translation(SMT),based on reviews of state-of-the-art technology.First the research team introduced the concept of paraphrases,and next a summarization of the latest progress utilizing paraphrasing technology in SMT was conducted.Finally,conclusions were drawn,data was compared and an analysis of the main issues of incorporating paraphrases into SMT,including translation model training,parameter tuning,input sentences rewriting and machine translation evaluation was performed.The results proved that there is an inherent connection between paraphrasing and SMT.The results also point out that the correctness and diversity of paraphrasing are the key issues to apply paraphrasing to SMT.It was highly noted that the improvement in the quality of paraphrasing resource,the establishment of a joint model of paraphrasing and machine translation and the new proposed approach to solve data sparseness are problems which need further study.