Annotating Event Anaphora: A Case Study

In recent years we have resgitered a renewed interest in event detection and temporal processing of text/discourse. TimeML (Pustejovsky et al., 2003a) has shed new lights on the notion of event and developed a new methodology for its annotation. On a parallel, works on anaphora resolution have developed a reliable methodology for the annotation and pointed out the core role of this phenomenon for the improvement of NLP systems. This paper tries to put together these two lines of research by describing a case study for the creation of an annotation scheme on event anaphora. We claim that this work could have consequences for the annotation of eventualities as proposed in TimeML and on the use of the tag and on the study of anaphora and its annotation. The annotation scheme and its guidelines have been developed on the basis of a coarse grained bottom up approach. In order to do this, we have performed a small sampling annotation which has highlighted shortcomings and open issues which need to be resolved.