Large-scale discriminative machine translation promises to further the state-of-the-art, but has failed to deliver convincing gains over current heuristic frequency count systems. We argue that a principle reason for this failure is not dealing with multiple, equivalent translations. We present a translation model which models derivations as a latent variable, in both training and decoding, and is fully discriminative and globally optimised. Results show that accounting for multiple derivations does indeed improve performance. Additionally, we show that regularisation is essential for maximum conditional likelihood models in order to avoid degenerate solutions. 1

We offer a simple, effective, and scalable method for statistical machine translation parameter tuning based on the pairwise approach to ranking (Herbrich et al., 1999). Unlike the popular MERT algorithm (Och, 2003), our pairwise ranking optimization (PRO) method is not limited to a handful of parameters and can easily handle systems with thousands of features. Moreover, unlike recent approaches built upon the MIRA algorithm of Crammer and Singer (2003) (Watanabe et al., 2007; Chiang et al., 2008b), PRO is easy to implement. It uses off-the-shelf linear binary classifier software and can be built on top of an existing MERT framework in a matter of hours. We establish PRO’s scalability and effectiveness by comparing it to MERT and MIRA and demonstrate parity on both phrase-based and syntax-based systems in a variety of language pairs, using large scale data scenarios. 1

We extend discriminative n-gram language modeling techniques originally proposed for automatic speech recognition to a statistical machine translation task. In this context, we propose a novel data selection method that leads to good models using a fraction of the training data. We carry out systematic experiments on several benchmark tests for Chinese to English translation using a hierarchical phrase-based machine translation system, and show that a discriminative language model significantly improves upon a state-of-the-art baseline. The experiments also highlight the benefits of our data selection method. 1

Translation systems are generally trained to optimize BLEU, but many alternative metrics are available. We explore how optimizing toward various automatic evaluation metrics (BLEU, METEOR, NIST, TER) affects the resulting model. We train a state-of-the-art MT system using MERT on many parameterizations of each metric and evaluate the resulting models on the other metrics and also using human judges. In accordance with popular wisdom, we find that it’s important to train on the same metric used in testing. However, we also find that training to a newer metric is only useful to the extent that the MT model’s structure and features allow it to take advantage of the metric. Contrasting with TER’s good correlation with human judgments, we show that people tend to prefer BLEU and NIST trained models to those trained on edit distance based metrics like TER or WER. Human preferences for METEOR trained models varies depending on the source language. Since using BLEU or NIST produces models that are more robust to evaluation by other metrics and perform well in human judgments, we conclude they are still the best choice for training. 1

This paper shows that discriminative reranking with an averaged perceptron model yields substantial improvements in realization quality with CCG. The paper confirms the utility of including language model log probabilities as features in the model, which prior work on discriminative training with log linear models for HPSG realization had called into question. The perceptron model allows the combination of multiple n-gram models to be optimized and then augmented with both syntactic features and discriminative n-gram features. The full model yields a stateof-the-art BLEU score of 0.8506 on Section 23 of the CCGbank, to our knowledge the best score reported to date using a reversible, corpus-engineered grammar. 1

State-of-the-art statistical machine translation (MT) systems have made significant progress towards producing user-acceptable translation output. However, there is still no efficient way for MT systems to inform users which words are likely translated correctly and how confident it is about the whole sentence. We propose a novel framework to predict wordlevel and sentence-level MT errors with a large number of novel features. Experimental results show that the MT error prediction accuracy is increased from 69.1 to 72.2 in F-score. The Pearson correlation between the proposed confidence measure and the human-targeted translation edit rate (HTER) is 0.6. Improvements between 0.4 and 0.9 TER reduction are obtained with the n-best list reranking task using the proposed confidence measure. Also, we present a visualization prototype of MT errors at the word and sentence levels with the objective to improve post-editor productivity. 1

This work proposes to adapt an existing general SMT model for the task of translating queries that are subsequently going to be used to retrieve information from a target language collection. In the scenario that we focus on access to the document collection itself is not available and changes to the IR model are not possible. We propose two ways to achieve the adaptation effect and both of them are aimed at tuning parameter weights on a set of parallel queries. The first approach is via a standard tuning procedure optimizing for BLEU score and the second one is via a reranking approach optimizing for MAP score. We also extend the second approach by using syntax-based features. Our experiments show improvements of 1-2.5 in terms of MAP score over the retrieval with the non-adapted translation. We show that these improvements are due both to the integration of the adaptation and syntax-features for the query translation task. 1

Although the scoring features of state-of-theart Phrase-Based Statistical Machine Translation (PB-SMT) models are weighted so as to optimise an objective function measuring translation quality, the estimation of the features themselves does not have any relation to such quality metrics. In this paper, we introduce a translation quality-based feature to PB-SMT in a bid to improve the translation quality of the system. Our feature is estimated by averaging the edit-distance between phrase pairs involved in the translation of oracle sentences, chosen by automatic evaluation metrics from the N-best outputs of a baseline system, and phrase pairs occurring in the N-best list. Using our method, we report a statistically significant 2.11 % relative improvement in BLEU score for the WMT 2009 Spanish-to-English translation task. We also report that using our method we can achieve statistically significant improvements over the baseline using many other MT evaluation metrics, and a substantial increase in speed and reduction in memory use (due to a reduction in phrase-table size of 87%) while maintaining significant gains in translation quality.

We propose a statistical sentence simplification system with log-linear models. In contrast to state-of-the-art methods that drive sentence simplification process by hand-written linguistic rules, our method used a margin-based discriminative learning algorithm operates on a feature set. The feature set is defined on statistics of surface form as well as syntactic and dependency structures of the sentences. A stack decoding algorithm is used which allows us to efficiently generate and search simplification hypotheses. Experimental results show that the simplified text produced by the proposed system reduces 1.7 Flesch-Kincaid grade level when compared with the original text. We will show that a comparison of a state-ofthe-art rule-based system (Heilman and Smith, 2010) to the proposed system demonstrates an improvement of 0.2, 0.6, and 4.5 points in ROUGE-2, ROUGE-4, and AveF10, respectively. 1

We present a scalable discriminative training framework for parsing-based statistical machine translation. Our framework exploits hypergraphs (or packed forests) to compactly encode exponentially many competing translations, and uses the perceptron algorithm to learn to discriminatively prefer the oracle-best tree in the hypergraph. To facilitate training, we present: (i) an oracle extraction algorithm to efficiently extract the oracle trees from a hypergraph that best match the reference translations; (ii) a hypergraph pruning algorithm that substantially reduces the disk space required for storing the hypergraphs without degrading the translation quality; and (iii) simple yet effective data and feature selection algorithms by which an equally good or better model is obtained using a fraction of the training data and features. We experimentally show that our approach is scalable and is able to improve over a full-scale state-of-the-art hierarchical machine translation system. 1