Parallel corpora are crucial for training SMT systems. However, for many language pairs they are available only in very limited quantities. For these language pairs a huge portion of phrases encountered at run-time will be unknown. We show how techniques from paraphrasing can be used to deal with these otherwise unknown source language phrases. Our results show that augmenting a stateof-the-art SMT system with paraphrases leads to significantly improved coverage and translation quality. For a training corpus with 10,000 sentence pairs we increase the coverage of unique test set unigrams from 48 % to 90%, with more than half of the newly covered items accurately translated, as opposed to none in current approaches. 1

Statistical machine translation (SMT) treats the translation of natural language as a machine learning problem. By examining many samples of human-produced translation, SMT algorithms automatically learn how to translate. SMT has made tremendous strides in less than two decades, and many popular techniques have only emerged within the last few years. This survey presents a tutorial overview of state-of-the-art SMT at the beginning of 2007. We begin with the context of the current research, and then move to a formal problem description and an overview of the four main subproblems: translational equivalence modeling, mathematical modeling, parameter estimation, and decoding. Along the way, we present a taxonomy of some different approaches within these areas. We conclude with an overview of evaluation and notes on future directions.

This paper describes a study of the patterns of translational equivalence exhibited by a variety of bitexts. The study found that the complexity of these patterns in every bitext was higher than suggested in the literature. These findings shed new light on why “syntactic ” constraints have not helped to improve statistical translation models, including finitestate phrase-based models, tree-to-string models, and tree-to-tree models. The paper also presents evidence that inversion transduction grammars cannot generate some translational equivalence relations, even in relatively simple real bitexts in syntactically similar languages with rigid word order. Instructions for replicating our experiments are at

Statistical machine translation relies heavily on the available training data. In some cases it is necessary to limit the amount of training data that can be created for or actually used by the systems. We introduce weighting schemes which allow us to sort sentences based on the frequency of unseen n-grams. A second approach uses TF-IDF to rank the sentences. After sorting we can select smaller training corpora and we are able to show that systems trained on much less training data achieve a very competitive performance compared to baseline systems using all available training data. 1.

Translation model size is growing at a pace that outstrips improvements in computing power, and this hinders research on many interesting models. We show how an algorithmic scaling technique can be used to easily handle very large models. Using this technique, we explore several large model variants and show an improvement 1.4 BLEU on the NIST 2006 Chinese-English task. This opens the door for work on a variety of models that are much less constrained by computational limitations.

We describe Joshua, an open source toolkit for statistical machine translation. Joshua implements all of the algorithms required for synchronous context free grammars (SCFGs): chart-parsing, n-gram language model integration, beamand cube-pruning, and k-best extraction. The toolkit also implements suffix-array grammar extraction and minimum error rate training. It uses parallel and distributed computing techniques for scalability. We demonstrate that the toolkit achieves state of the art translation performance on the WMT09 French-English translation task. 1

This work summarizes a comparison between two approaches to Statistical Machine Translation (SMT), namely Ngram-based and Phrase-based SMT. In both approaches, the translation process is based on bilingual units related by word-to-word alignments (pairs of source and target words), while the main differences are based on the extraction process of these units and the statistical modeling of the translation context. The study has been carried out on two different translation tasks (in terms of translation difficulty and amount of available training data), and allowing for distortion (reordering) in the decoding process. Thus it extends a previous work were both approaches were compared under monotone conditions. We finally report comparative results in terms of translation accuracy, computation time and memory size. Results show how the ngram-based approach outperforms the phrase-based approach by achieving similar accuracy scores in less computational time and with less memory needs. 1.

A principal weakness of conventional (i.e., non-hierarchical) phrase-based statistical machine translation is that it can only exploit continuous phrases. In this paper, we extend phrase-based decoding to allow both source and target phrasal discontinuities, which provide better generalization on unseen data and yield significant improvements to a standard phrase-based system (Moses). More interestingly, our discontinuous phrasebased system also outperforms a state-of-the-art hierarchical system (Joshua) by a very significant margin (+1.03 BLEU on average on five Chinese-English NIST test sets), even though both Joshua and our system support discontinuous phrases. Since the key difference between these two systems is that ours is not hierarchical—i.e., our system uses a string-based decoder instead of CKY, and it imposes no hard hierarchical reordering constraints during training and decoding—this paper sets out to challenge the commonly held belief that the tree-based parameterization of systems such as Hiero and Joshua is crucial to their good performance against Moses. 1

We describe a scalable decoder for parsing-based machine translation. The decoder is written in Java and implements all the essential algorithms described in (Chiang, 2007) and (Li and Khudanpur, 2008b): chart-parsing, n-gram language model integration, beam- and cube-pruning, and k-best extraction. Additionally, parallel and distributed computing techniques are exploited to make it scalable. We demonstrate experimentally that our decoder is more than 30 times faster than a baseline decoder written in Python.

Accurately translating multiword expressions is important to obtain good performance in machine translation, crosslanguage information retrieval, and other multilingual tasks in human language technology. Existing approaches to inducing translation equivalents of multiword units have focused on agglomerating individual words or on aligning words in a statistical machine translation system. We present a different approach based upon information theoretic heuristics and the exact counting of frequencies of occurrence of multiword strings in aligned parallel corpora. We are applying a technique introduced by Yamamoto and Church that uses suffix arrays and longest common prefix arrays. Evaluation of the method in multiple language pairs was performed using bilingual lexicons of domainspecific terminology as a gold standard. We found that performance of 50-70%, as measured by mean reciprocal rank, can be obtained for terms that occur more than 10 or so times. 1