An algorithm for zero pronoun resolution in Bulgarian

Abstract

This paper presents an algorithm for identifying the noun phrase antecedents of zero pronouns in Bulgarian. The algorithm applies to syntactic representation in XML format, generated by context-free grammar parser. Like the parser, the algorithm is implemented in Java. The author has tested it on Z-corpora, specially created for the investigation of zero pronouns in Bulgarian. The corpora are manually annotated and contain 1029 zero pronouns occurrences. The algorithm has 84% critical success rate. The algorithm is compared to another approach, which has been proposed in literature. The author's algorithm achieves 3% higher critical success rate.