Improving Anaphora Resolution performance in Italian and English using Search Engines
Fabio Celli · 2010
In this paper I describe an euristic method for improving anaphora resolution by classifying as co-referring the pairs in a high co-occurrence with taxonomic relations (Is-A) and as non-coreferring the pairs in a high co-occurrence with associative relations (Part-Of). The experiments re-ported here show that it is posible to classify correctly 65 % of the errors of and anaphora resolution system by using simple pattern-based relation extraction from web. 1