Learning Multiple-Nonterminal Synchronous Grammars for Statistical Machine Translation

Recent work in machine translation has evolved from the traditional word and phrase based models to include hierarchical phrase based and syntax-based models. These advances are motivated by the desire to integrate richer knowledge within the translation process to explicitly address limitations of the purely lexical phrasebased model. Generalized phrases as discussed in (Chiang, 2005) attempt to directly address the limitiations of purely lexical phrases, and have shown significant improvements in translation quality by introducing constructs for sub-phrase representation. However, generalizations are represented by a single sub-phrase category (and a glue rule for serial combination), providing the ability (and risk) of inserting any available sub-phrase into a larger phrase. The first contribution of this dissertation work is syntax-augmented machine translation (SAMT), an extension to Chiang’s model that provides multiple generalization types based on the phrase-structure parse trees of the training target sentences. We report improvements over strong phrase-based as well as hierarchical phrase-based baselines for French-to-English, Chinese-to-English, and Urdu-toEnglish. Syntax-based models such as SAMT typically rely on word-alignments and parse trees of the training sentence pairs, which are assumed to be correct. In reality, these alignments and parses are not human-generated, but instead result from the most probable configuration of a stochastic model. We provide a method to induce grammars over hidden alignments and parses, approximated from N -best lists. We present results showing improvements for hierarchical phrase-based MT as well as SAMT when using the widened pipeline. The SAMT model presupposes the availability of phrase-structure parse trees for the target training sentences. However, syntactic parsers are only available for a limited set of languages. We propose a labeling approach that is based merely on part-of-speech analysis of the source or target language (or even both). When using English POS tags in our labeling approach we achieve improvements in translation quality over Chiang’s hierarchical phrase-based MT model. We propose the application of the model to automatically learned word tags as future dissertation work. We further propose to induce a multi-nonterminal grammar from training data without any linguistic annotations based on a generative model over initial rules whose labels are latent variables. Our algorithm’s underlying model assigns class labels to phrase pairs. These labeled phrase pairs are then used as initial rules fed to the rule extraction algorithm of Chiang (2005), resulting in a synchronous gram-