Proceedings from the First Bergen Workshop on Anaphora Resolution (War I)

This book presents recent research in the fields of anaphora resolution and co-reference. The contributions cover a wide range of approaches, most of which involve computational models or machine- learning techniques. The basic problems of referent analysis of text are also treated. Examples of applications include answer extraction for Q&A systems, and processing ontological information. One experiment describes the development of pronominal anaphora resolution for Tamil, using a centering approach compared to a statistically induced model. Novel research is presented on how to improve the quality of co-reference chains using a spell correction approach. The development of a graphical web interface for presenting and editing co-reference chains is described with links to an online demonstrator. Finally, an extensive annotation guideline for coding various anaphor-antecedent pairs is given. The guideline is a good starting point for any project aiming at annotating large corpora for co-reference, and it is an essential resource to assure good agreement between annotators. The work represents international efforts, with contributors from Sweden, Norway, the Netherlands, Belgium and India. A broad view of the role of anaphora and co-reference in Language Technology and Computational Linguistics is presented. Insights and ideas for further research are presented for those interested in defining their own project within this important, but difficult, field. The book will also be useful for the beginner in the field, as it presents a wide selection of research in the area.