The Person Cros Document Coreference sys- tems depend on the context for making deci- sions on the posible coreferences betwen person name mentions. The amount of context required is a parameter that varies from cor- pora to corpora, which makes it dificult for usual disambiguation methods. In this paper we show that the amount of context required can be dynamicaly controled on the basis of the prior probabilities of coreference and we present a new statistical model for the compu- tation of these probabilities. The experiment we caried on a news corpus proves that the prior probabilities of coreference are an impor- tant factor for maintaining a god balance be- twen precision and recal for cros document coreference systems.

Click Here to View the Paper In XML Format