Staord Street, Wolverhampton,

Abstract. This paper describes a new, advanced and completely re-vamped version of Mitkov's knowledge-poor approach to pronoun res-olution [21]. In contrast to most anaphora resolution approaches, the new system, referred to as MARS, operates in fully automatic mode. It bene ts from purpose-built programs for identifying occurrences of non-nominal anaphora (including pleonastic pronouns) and for recognition of animacy, and employs genetic algorithms to achieve optimal perfor-mance. The paper features extensive evaluation and discusses important evaluation issues in anaphora resolution. 1 The original approach Mitkov's approach to anaphora resolution [21] avoids complex syntactic, seman-tic and discourse analysis relying on a list of preferences known as antecedent indicators. The approach operates as follows: it works on texts rst processed by a part-of-speech tagger and a noun phrase (NP) extractor, locates NPs which precede the anaphor within a distance of 2 sentences, checks them for gender and number agreement with the anaphor and then applies indicators to the re-maining candidates that assign positive or negative scores to them (-1, 0, 1 or 2). The NP with the highest composite score is proposed as antecedent 1 The antecedent indicator