James Clarke and Mirella lapata win Best paper award

James Clarke and Mirella Lapata have won the best paper award at the 2007 Joint Conference on Empirical Methods for Natural Language Processing, one of the most prestigious conferences in the field of natural language processing.

Their paper, Modelling Compression with Discourse Constraints, presents a new method for compressing documents. The task is important for a number of real world applications ranging from summarisation to information retrieval, and displaying text on small screens such as PDAs or mobile phones. The proposed solution makes it possible to combine a variety of knowledge sources in a principled manner using integer linear programming, a technique for solving discrete optimization problems that has proved valuable for a variety of problems in planning, routing, and scheduling. A large scale evaluation of the compression model with real users shows that it outperforms other more conventional approaches.

The paper will be presented at EMNLP Conference on 28 June 2007. View abstract.