Lexical Chain Based Cohesion Models for Document-Level Statistical Machine Translation

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 1563–1573, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Lexical Chain Based Cohesion Models for Document-Level Statistical Machine Translation Deyi Xiong1, Yang Ding2, Min Zhang1∗ and Chew Lim Tan2 1School of Computer Science and Technology, Soochow University, Suzhou, China 215006 {dyxiong, minzhang}@suda.edu.cn 2School of Computing, National University of Singapore, Singapore 117417 {a0082379, tancl}@comp.nus.edu.sg Abstract Lexical chains provide a representation of the lexical cohesion structure of a text. In this pa- per, we propose two lexical chain based co- hesion models to incorporate lexical cohesion into document-level statistical machine trans- lation: 1) a count cohesion model that rewards a hypothesis whenever a chain word occurs in the hypothesis, 2) and a probability cohesion model that further takes chain word transla- tion probabilities into account. We compute lexical chains for each source document to be translated and generate target lexical chains based on the computed source chains via max- imum entropy classifiers. We then use the generated target chains to provide constraints for word selection in document-level machine translation through the two proposed lexical chain based cohesion models. We verify the effectiveness of the two models using a hier- archical phrase-based translation system. Ex- periments on large-scale training data show that they can substantially improve translation quality in terms of BLEU and that the prob- ability cohesion model outperforms previous models based on lexical cohesion devices. 1 Introduction Given a source document, traditionally most statisti- cal machine translation (SMT) systems translate the document sentence by sentence. In such a transla- tion scheme, sentences are translated independent of any other sentences. However, a text is normally written cohesively, in which sentences are connected ∗Corresponding author to each other via syntactic and lexical devices. This linguistic phenomenon is called as textual cohesion (Halliday and Hasan, 1976). Cohesion is a surface-level property of well- formed texts. It deals with five categories of rela- tionships between text units, namely co-reference, ellipsis, substitution, conjunction and lexical cohe- sion that is realized via semantically related words. The former four cohesion relations can be grouped as grammatical cohesion. Generally speaking, grammatical cohesion is less common and harder to identify than lexical cohesion (Barzilay and El- hadad, 1997). As most SMT systems translate a text in a sentence-by-sentence fashion, they tend to build less lexical cohesion than human translators (Wong and Kit, 2012). We therefore study lexical cohesion for document-level translation. We use lexical chains (Morris and Hirst, 1991) to capture lexical cohe- sion in a text. Lexical chains are connected graphs that represent the lexical cohesion structure of a text. They have been successfully used for information retrieval (Stairmand, 1996), document summariza- tion (Barzilay and Elhadad, 1997) and so on. In this paper, we investigate how lexical chains can be used to incorporate lexical cohesion into document-level translation. Our basic assumption is that the lexical chains of a target document are direct correspondences of the lexical chains of its counterpart source document. This …

Content truncated to 500 words