Many probabilistic models for natural language are now written in terms of hierarchical tree structure. Tree-based modeling still lacks many of the standard tools taken for granted in (finite-state) string-based modeling. The theory of tree transducer automata provides a possible framework to draw on, as it has been worked out in an extensive literature. We motivate the use of tree transducers for natural language and address the training problem for probabilistic tree-totree and tree-to-string transducers.

We present Minimum Bayes-Risk (MBR) decoding for statistical machine translation. This statistical approach aims to minimize expected loss of translation errors under loss functions that measure translation performance. We describe a hierarchy of loss functions that incorporate different levels of linguistic information from word strings, word-to-word alignments from an MT system, and syntactic structure from parse-trees of source and target language sentences. We report the performance of the MBR decoders on a Chinese-to-English translation task. Our results show that MBR decoding can be used to tune statistical MT performance for specific loss functions. 1

A syntax-directed translator first parses the source-language input into a parsetree, and then recursively converts the tree into a string in the target-language. We model this conversion by an extended treeto-string transducer that have multi-level trees on the source-side, which gives our system more expressive power and flexibility. We also define a direct probability model and use a linear-time dynamic programming algorithm to search for the best derivation. The model is then extended to the general log-linear framework in order to rescore with other features like n-gram language models. We devise a simple-yet-effective algorithm to generate non-duplicate k-best translations for n-gram rescoring. Initial experimental results on English-to-Chinese translation are presented. 1

This paper presents novel approaches to reordering in phrase-based statistical machine translation. We perform consistent reordering of source sentences in training and estimate a statistical translation model. Using this model, we follow a phrase-based monotonic machine translation approach, for which we develop an efficient and flexible reordering framework that allows to easily introduce different reordering constraints. In translation, we apply source sentence reordering on word level and use a reordering automaton as input. We show how to compute reordering automata on-demand using IBM or ITG constraints, and also introduce two new types of reordering constraints. We further add weights to the reordering automata. We present detailed experimental results and show that reordering significantly improves translation quality. 1

This paper investigates some computational problems associated with probabilistic translation models that have recently been adopted in the literature on machine translation. These models can be viewed as pairs of probabilistic contextfree grammars working in a ‘synchronous’ way. Two hardness results for the class NP are reported, along with an exponential time lower-bound for certain classes of algorithms that are currently used in the literature. 1

We study automata for capturing transformations employed by practical natural language processing systems, such as those that translate between human languages. For several variations of finite-state string and tree transducers, we ask formal questions about expressiveness, modularity, teachability, and generalization.

We generalise existing forward and backward bisimulation minimisation algorithms for tree automata to weighted tree automata. The obtained algorithms work for all semirings and retain the time complexity of their unweighted variants for all additively cancellative semirings. On all other semirings the time complexity is slightly higher (linear instead of logarithmic in the number of states). We discuss implementations of these algorithms on a typical task in natural language processing.

In this paper, we present a novel training method for a localized phrase-based prediction model for statistical machine translation (SMT). The model predicts block neighbors to carry out a phrasebased translation that explicitly handles local phrase re-ordering. We use a maximum likelihood criterion to train a log-linear block bigram model which uses real-valued features (e.g. a language model score) as well as binary features based on the block identities themselves (e.g. block bigram features). The model training relies on an efficient enumeration of local block neighbors in parallel training data. A novel stochastic gradient descent (SGD) training algorithm is presented that can easily handle millions of features. Moreover, when viewing SMT as a block generation process, it becomes quite similar to sequential natural language annotation problems such as part-of-speech tagging, phrase chunking, or shallow parsing. The novel approach is successfully tested on a standard Arabic-English translation task using two different phrase re-ordering models: a block orientation model and a phrase-distortion model. Categories and Subject Descriptors: I.2.7 [Artificial Intelligence]: Natural Language Processing—statistical machine translation; G.3 [Probability and Statistics]: Statistical computing— stochastic gradient descent

Language Weaver Inc. Designers of statistical machine translation (SMT) systems have begun to employ tree-structured translation models. Systems involving tree-structured translation models tend to be complex. This article aims to reduce the conceptual complexity of such systems, in order to make them easier to design, implement, debug, use, study, understand, explain, modify, and improve. In parsing algorithm with five functional parameters: a logic, a grammar, a semiring, a search strategy, and a termination condition. The article then shows that all the common algorithms that revolve around tree-structured translation models, including hierarchical alignment, inference for parameter estimation, translation, and structured evaluation, can be derived by generalizing two of these parameters — the grammar and the logic. The article culminates with a recipe for using such generalized parsers to train, apply, and evaluate an SMT system that is driven by tree-structured translation models. 1.

Phrase alignment is the task that requires the constituent phrases of two halves of a bitext to be aligned. In order to align phrases, one must discover them first and this article presents a method of aligning phrases that are discovered automatically. Here, the notion of a ‘phrase ’ will be understood as being given by a subtree of a dependency-like structure of a sentence called linkage. To discover phrases, we will make use of two distinct, language independent methods: the IBM-1 model (Brown et al., 1993) adapted to detect linkages and Constrained Lexical Attraction Models (Ion & Barbu Mititelu, 2006). The methods will be combined and the resulted model will be used to annotate the bitext. The accuracy of phrase alignment will be evaluated by obtaining word alignments from link alignments and then by checking the F-measure of the latter word aligner. 1.