Traditionally, statistical machine translation systems have relied on parallel bi-lingual data to train a translation model. While bi-lingual parallel data are expensive to generate, monolingual data are relatively common. Yet monolingual data have been under-utilized, having been used primarily for training a language model in the target language. This paper describes a novel method for utilizing monolingual target data to improve the performance of a statistical machine translation system on news stories. The method exploits the existence of comparable text—multiple texts in the target language that discuss the same or similar stories as found in the source language document. For every source document that is to be translated, a large monolingual data set in the target language is searched for documents that might be comparable to the source documents. These documents are then used to adapt the MT system to increase the probability of generating texts that resemble the comparable document. Experimental results obtained by adapting both the language and translation models show substantial gains over the baseline system. 1

The translation model of statistical machine translation systems is trained on parallel data coming from various sources and domains. These corpora are usually concatenated, word alignments are calculated and phrases are extracted. This means that the corpora are not weighted according to their importance to the domain of the translation task. This is in contrast to the training of the language model for which well known techniques are used to weight the various sources of texts. On a smaller granularity, the automatic calculated word alignments differ in quality. This is usually not considered when extracting phrases either. In this paper we propose a method to automatically weight the different corpora and alignments. This is achieved with a resampling technique. We report experimental results for a small (IWSLT) and large (NIST) Arabic/English translation tasks. In both cases, significant improvements in the BLEU score were observed. 1

We show that unseen words account for a large part of the translation error when moving to new domains. Using an extension of a recent approach to mining translations from comparable corpora (Haghighi et al., 2008), we are able to find translations for otherwise OOV terms. We show several approaches to integrating such translations into a phrasebased translation system, yielding consistent improvements in translations quality (between 0.5 and 1.5 Bleu points) on four domains and two language pairs. 1

Statistical machine translation systems are usually trained on a large amount of bilingual sentence pairs and translate one sentence at a time, ignoring document-level information. In this paper, we propose a cache-based approach to document-level translation. Since caches mainly depend on relevant data to supervise subsequent decisions, it is critical to fill the caches with highly-relevant data of a reasonable size. In this paper, we present three kinds of caches to store relevant document-level information: 1) a dynamic cache, which stores bilingual phrase pairs from the best translation hypotheses of previous sentences in the test document; 2) a static cache, which stores relevant bilingual phrase pairs extracted from similar bilingual document pairs (i.e. source documents similar to the test document and their corresponding target documents) in the training parallel corpus; 3) a topic cache, which stores the target-side topic words related with the test document in the source-side. In particular, three new features are designed to explore various kinds of document-level information in above three kinds of caches. Evaluation shows the effectiveness of our cache-based approach to document-level translation with the performance improvement of 0.81 in BLUE score over Moses. Especially, detailed analysis and discussion are presented to give new insights to document-level translation. 1

Statistical machine translation (SMT) systems use statistical learning methods to learn how to translate from large amounts of parallel training data. Unfortunately, SMT systems are tuned to the domain of the training data and need to be adapted before they can be used to translate data in a different domain. First, we consider a semi-supervised technique to perform model adaptation. We explore new feature extraction techniques, feature combinations and their effects on performance. In addition, we introduce an unsupervised variant of Minimum Error Rate Training (MERT), which can be used to tune the SMT model parameters. We do this by using another SMT model that translates in the reverse direction. We apply this variant of MERT to the model adaptation task. Both of the techniques we explore in this thesis produce promising results in exhaustive experiments we performed for translation from French to English in different domains.

In the past decade we have seen an amazing revival of machine translation (MT) as the major field of research in computational linguistics. Many reasons can be mentioned to explain this phenomenon: Globalization and the success of the Internet may be one of them forcing companies and individuals to adapt to a multilingual

Statistical machine translation relies heavily on available parallel corpora, but SMT may not have the ability or intelligence to make full use of the training set. Instead of collecting more and more parallel training corpora, this paper aims to improve SMT performance by exploiting the full potential of existing parallel corpora. We first identify literally translated sentence pairs via lexical and grammatical compatibility, and then use these data to train SMT models. One experiment indicates that larger training corpora do not always lead to higher decoding performance when the added data

The Cunei Machine Translation Platform is an open-source MT system designed to model instances of translation. One of the challenges to this approach is effective training. We describe two techniques that improve the training procedure and allow us to leverage the strengths of instance-based modeling. First, during training we approximate our model with a second-order Taylor series. Second, we discount models based on the magnitude of their approximation. By reducing error in training, our model now consistently outperforms the standard SMT model with gains ranging from 0.51 to 3.77 BLEU on German-English and Czech-English test sets. 1

We describe our system for the news commentary translation task of WMT 2011. The submitted run for the French-English direction is a combination of two MOSES-based systems developed at LIG and LIA laboratories. We report experiments to improve over the standard phrase-based model using statistical post-edition, information retrieval methods to subsample out-of-domain parallel corpora and ROVER to combine n-best list of hypotheses output by different systems. 1

This paper addresses the problem of dynamic model parameter selection for loglinear model based statistical machine translation (SMT) systems. In this work, we propose a principled method for this task by transforming it to a test data dependent development set selection problem. We present two algorithms for automatic development set construction, and evaluated our method on several NIST data sets for the Chinese-English translation task. Experimental results show that our method can effectively adapt log-linear model parameters to different test data, and consistently achieves good translation performance compared with conventional methods that use a fixed model parameter setting across different data sets. 1