HLT - NAACL - 2006 Computationally Hard Problems and Joint Inference in Speech and Language Processing

Recent work on ranking, sampling and other approximate solutions to natural language processing problems indicate that researchers are coming back to the hard problems in speech and text, for which efficient algorithms are not known to exist. In addition, there has been increasing interest in moving away from systems that make chains of local decisions independently, and instead toward systems that make multiple decisions jointly using global information. The goal of this workshop is to bring together researchers working on NLP problems whose solutions are computationally hard—whether because the problem is not well modeled by only local features, or because the problem is best solved in a joint, rather than pipelined, manner. We are grateful to the program committee for providing thoughtful and helpful reviews of the submitted papers. We also thank our invited speakers, we thank the organizers of the main HLT/NAACL 2006 conference, without which this workshop would not be possible. Abstract A syntax-directed translator first parses the source-language input into a parse-tree, and then recursively converts the tree into a string in the target-language. We model this conversion by an extended tree-to-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.