Discourse Deixis and Coreference: Evidence from AnCora
Marta Vilar Recasens · DSpace repository (University of Tartu) · 2008
Few empirical studies have been conducted on discourse deixis, and no such study exists for Catalan or Spanish. This paper presents an em-pirical analysis of 200 000 words from the An-Cora corpora annotated with discourse deixis. It returns to and tests assumptions previously made, laying out the linguistic problems we still need to account for. To this end, proposals are put forward with regard to (i) the detec-tion of abstract anaphors, and (ii) the way their antecedents should be understood, drawing on the theory of underspecification. The quantita-tive and qualitative corpus analysis casts light on ways of improving the performance of coref-erence resolution systems by shifting the focus from the delimitation of antecedents to the de-tection of abstract anaphors. 1