Authors: Sheetal S. Sonawane; Parag A. Kulkarni
Addresses: Department of Computer Engineering, College of Engineering, Pune 411005, Maharashtra, India ' Department of Computer Engineering, College of Engineering, Pune 411005, Maharashtra, India
Abstract: Co-reference resolution is a method of finding an association of feature terms or mentions in the discourse that refers to the same entity. Existing graph based methods for co-reference resolution does not consider context based relationship between feature terms. Edge weight in existing graph based method is calculated using lexical, syntactic and semantic features. To address this issue, context based co-reference resolution using graph model is presented in this paper. Edge weight is calculated using intelligent set of feature functions. The research is done in two directions: 1) Context building of a term or feature using Wikipedia for construction of semantic features towards performance improvement. 2) Resolving co-reference using graph clustering approach. The experimental results on Reuter's, UMIREC and ACE 2002 dataset show that our approach outperforms to that of state of the art model.
Keywords: co-reference resolution; graph clustering; graph models; information retrieval; natural language processing; NLP; text documents; context based relationships; text features; text terms; semantic features.
International Journal of Knowledge Engineering and Data Mining, 2016 Vol.4 No.1, pp.1 - 17
Received: 13 Aug 2015
Accepted: 11 Mar 2016
Published online: 06 Feb 2017 *