Selected Publications

• Christopher Dyer, Smaranda Muresan and Philip Resnik (2008). Generalizing Word Lattice Translation. Proceedings of the Association for Computational Linguistics, ACL 2008 .
  
  Abstract: Word lattice decoding has proven useful in spoken language translation; we argue that it provides a compelling model for translation of text genres, as well. Additionally, we show that prior work in translating lattices using finite state techniques can be naturally extended to more expressive synchronous context-free grammar-based models. In the process, we resolve a significant complication that lattice representations introduce in reordering models. Our experiments evaluating the approach demonstrate substantial gains for Chinese-English and Arabic-English translation.

• Smaranda Muresan and Owen Rambow (2007). Grammar Approximation by Representative Sublanguage: A New Model for Language Learning . In Proceedings of the Association for Computational Linguistics, ACL 2007, Prague, Czech Republic.
  
  Abstract: We propose a new language learning model that learns a syntactic-semantic grammar from a small number of natural language strings annotated with their semantics, along with basic assumptions about natural language syntax. We show that the search space for grammar induction is a complete grammar lattice, which guarantees the uniqueness of the learned grammar.
  (see also --- Smaranda Muresan 2005. Parsing Preserving Techniques in Grammar Induction. Technical Report CUCS-032-05, Columbia University, New York, NY. )

• Smaranda Muresan (2006). Learning Constraint-based Grammars from Representative Examples: Theory and Applications . PhD Dissertation, Columbia University, New York, 2006
  
  Short Abstract Theoretical Part : This dissertation defines a new type of constraint-based grammars, Lexicalized Well-Founded Grammars , which establish a direct link between natural language expressions and ontologies, facilitating a direct mapping between text and knowledge (Chapter 3). Semantic composition and semantic interpretation are defined as constraints at the grammar rule level (Chapter 4). I introduced a new grammar learning model from representative data Grammar Approximation by Representative Sublanguage , for learning LWFGs (Chapter 5). I have proven that the search space for grammar induction is a complete grammar lattice, and I have provided polynomial algorithms for grammar induction and proved that they are correct (Chapter 5). Applicative Part: I have embedded all these theoretical concepts in a practical, implemented system for grammar learning (Chapter 7). I have provided qualitative evaluations that cover the following issues: coverage of diverse and complex linguistic phenomena (Chapter 6); terminological knowledge acquisition from natural language definitions in the medical domain, as well as NL-query of this knowledge base (Chapter 8).