In adding syntax to statistical MT, there is a tradeoff between taking advantage of linguistic analysis, versus allowing the model to exploit linguistically unmotivated mappings learned from parallel training data. A number of previous efforts have tackled this tradeoff by starting with a commitment to linguistically motivated analyses and then finding appropriate ways to soften that commitment. We present an approach that explores the tradeoff from the other direction, starting with a context-free translation model learned directly from aligned parallel text, and then adding soft constituent-level constraints based on parses of the source language. We obtain substantial improvements in performance for translation from Chinese and Arabic to English. 1

Parsing and translating natural languages can be viewed as problems of predicting tree structures. For machine learning approaches to these predictions, the diversity and high dimensionality of the structures involved mandate very large training sets. This paper presents a purely discriminative learning method that scales up well to problems of this size. Its accuracy was at least as good as other comparable methods on a standard parsing task. To our knowledge, it is the first purely discriminative learning algorithm for translation with treestructured models. Unlike other popular methods, this method does not require a great deal of feature engineering a priori, because it performs feature selection over a compound feature space as it learns. Experiments demonstrate the method’s versatility, accuracy, and efficiency. Relevant software is freely available at

This paper proposes a statistical, treeto-tree model for producing translations. Two main contributions are as follows: (1) a method for the extraction of syntactic structures with alignment information from a parallel corpus of translations, and (2) use of a discriminative, featurebased model for prediction of these targetlanguage syntactic structures—which we call aligned extended projections, or AEPs. An evaluation of the method on translation from German to English shows similar performance to the phrase-based model of Koehn et al. (2003). 1

We present a simple history-based model for sentence generation from LFG f-structures, which improves on the accuracy of previous models by breaking down PCFG independence assumptions so that more f-structure conditioning context is used in the prediction of grammar rule expansions. In addition, we present work on experiments with named entities and other multi-word units, showing a statistically significant improvement of generation accuracy. Tested on section 23 of the Penn Wall Street Journal Treebank, the techniques described in this paper improve BLEU scores from 66.52 to 68.82, and coverage from 98.18 % to 99.96%. 1

The paper discusses the transferring rules of the output from a dependency parser for Bulgarian into RMRS analyses. This task is required by the machine translation compatibility between Bulgarian and English resources. Since the Bulgarian HPSG grammar is still being developed, a repairing mechanism has been envisaged by parsing the Bulgarian data with the Malt Dependency Parser, and then retrieving RMRS analyses by exploring the linguistic knowledge within BulTreeBank-DP. 1

One style of Multi-Engine Machine Translation architecture involves choosing the best of a set of outputs from different systems. Choosing the best translation from an arbitrary set, even in the presence of human references, is a difficult problem; it may prove better to look at mechanisms for making such choices in more restricted contexts. In this paper we take a classificationbased approach to choosing between candidates from syntactically informed translations. The idea is that using multiple parsers as part of a classifier could help detect syntactic problems in this context that lead to bad translations; these problems could be detected on either the source side—perhaps sentences with difficult or incorrect parses could lead to bad translations—or on the target side—perhaps the output quality could be measured in a more syntactically informed way, looking for syntactic abnormalities. We show that there is no evidence that the source side information is useful. However, a target-side classifier, when used to identify particularly bad translation candidates, can lead to significant improvements in Bleu score. Improvements are even greater when combined with existing language and alignment model approaches. c ○ 2008. Licensed under the Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported license

Tree-to-string translation rules are widely used in linguistically syntax-based statistical machine translation systems. In this paper, we propose to use deep syntactic information for obtaining fine-grained translation rules. A head-driven phrase structure grammar (HPSG) parser is used to obtain the deep syntactic information, which includes a fine-grained description of the syntactic property and a semantic representation of a sentence. We extract fine-grained rules from aligned HPSG tree/forest-string pairs and use them in our tree-to-string and string-to-tree systems. Extensive experiments on largescale bidirectional Japanese-English translations testified the effectiveness of our approach. 1

In this paper, we describe a statistical deep syntactic transfer decoder that is trained fully automatically on parsed bilingual corpora. Deep syntactic transfer rules are induced automatically from the f-structures of a LFG parsed bitext corpus by automatically aligning local f-structures, and inducing all rules consistent with the node alignment. The transfer decoder outputs the n-best TL f-structures given a SL f-structure as input by applying large numbers of transfer rules and searching for the best output using a log-linear model to combine feature scores. The decoder includes a fully integrated dependency-based tri-gram language model. We include an experimental evaluation of the decoder using different parsing disambiguation resources for the German data to provide a comparison of how the system performs with different German training and test parses. 1