Pronoun Disambiguation Problem

In this paper we present current approaches for natural language understanding (NLU) applied on pronoun disambiguation problem (PDP). NLU is one of most fascinating fields within natural language processing (NLP) that deals with extraction of “deeper” meaning within text written in natural language. It is easy to demonstrate difficulties of understanding even simpler sentences like “Scientists have cloned monkeys.” unless we have a way of understanding that scientists don't necessarily possess cloned monkeys but have made a scientific breakthrough. Natural language is full of such ambiguities, incompleteness or vagueness and relies heavily on “common sense” of the reader. One of the most representative problem within NLU is PDP, where task is to assign a pronoun to one of previously mentioned nouns. Classical example of this problem is sentence “The city councilmen refused the demonstrators a permit because they feared violence.” taken from Winograd's test collection, where pronoun “they” refers to “councilmen” and not to “demonstrators” although both interpretations are syntactically and grammatically correct. This problem is often trivial for humans but very hard for known algorithms. Winograd, in his thesis, assumes that it is impossible to solve this problem without near-human understanding of the language and world in general, basically solving the problem of general artificial intelligence.