Improving Translation Fluency with Search-Based Decoding and a Monolingual Statistical Machine Translation Model for Automatic Post-Editing

Improving Translation Fluency with Search-Based Decoding and a Monolingual Statistical Machine Translation Model for Automatic Post-Editing Jing-Shin Chang Sheng-Sian Lin Department of Computer Science & Information Engineering Department of Computer Science & Information Engineering National Chi Nan University National Chi Nan University 1, Univ. Road, Puli, Nantou 545, TAIWAN 1, Univ. Road, Puli, Nantou 545, TAIWAN jshin@csie.ncnu.edu.tw s94321509@ncnu.edu.tw Abstract The BLEU scores and translation fluency for the current state-of-the-art SMT systems based on IBM models are still too low for publication purposes. The major issue is that stochastically generated sentences hypotheses, produced through a stack decoding process, may not strictly follow the natural target language grammar, since the decoding process is directed by a highly simplified translation model and n-gram language model, and a large number of noisy phrase pairs may introduce significant search errors. This paper proposes a statistical post-editing (SPE) model, based on a special monolingual SMT paradigm, to “translate” disfluent sentences into fluent sentences. However, instead of conducting a stack decoding process, the sentence hypotheses are searched from fluent target sentences in a large target language corpus or on the Web to ensure fluency. Phrase-based local editing, if necessary, is then applied to correct weakest phrase alignments between the disfluent and searched hypotheses using fluent target language phrases; such phrases are segmented from a large target language corpus with a global optimization criterion to maximize the likelihood of the training sentences, instead of using noisy phrases combined from bilingually word- aligned pairs. With such search-based decoding, the absolute BLEU scores are much higher than automatic post editing systems that conduct a classical SMT decoding process. We are also able to fully correct a significant number of disfluent sentences into completely fluent versions. The BLEU scores are significantly improved. The evaluation shows that on average 46% of translation errors can be fully recovered, and the BLEU score can be improved by about 26%. Keywords: Translation Fluency, Fluency-Based Decoding, Search-Based Decoding, Statistical Machine Translation, Automatic Post-Editing 1 Introduction and Motivation 1.1 Fluency Problems with Statistical Machine Translations Translation fluency of Machine Translation systems is a serious issue in the current SMT research works. With the research efforts for the past tens of years, the performances are still far from satisfactory. In translating English to Chinese, for instance, the BLEU scores [16] range only between 0.21 and 0.29 [22, 5, 17], depending on test sets and numbers of reference translations. Such translation quality is extremely disfluent for human readers. We therefore propose a statistical post-editing (SPE) model, based on a special monolingual SMT framework, for improving the fluency and adequacy of translated sentences. 195 The classical IBM SMT models [1, 2] formulate the translation problem of a source sentence F as finding the best translation E* from some stack decoded hypotheses, E, such that:       E E F EF E E E Pr | maxPr arg | argmaxPr *    (1) :target sentence where :source sentence FE and     Pr …

Content truncated to 500 words