EXmatcher: Combining Features Based on Reference Strings and Segments to Enhance Citation Matching

• DOI: 10.48550/arxiv.1906.04484
• Publication Date: 2019-01-01
• Authors: Behnam Ghavimi; Wolfgang Otto; Philipp Mayr
• Keywords: FOS: Computer and information sciences; 0202 electrical engineering, electronic engineering, information engineering; Computer Science - Digital Libraries; Digital Libraries (cs.DL); 02 engineering and technology; Information Retrieval (cs.IR); Computer Science - Information Retrieval

Citation matching is a challenging task due to different problems such as the variety of citation styles, mistakes in reference strings and quality identified segments. The classic configuration used this paper combination blocking technique binary classifier. Three possible inputs (reference strings, segments segments) were tested find most efficient strategy for matching. In classification step, we describe effect which probabilities can have Our evaluation on manually curated gold standard showed that input data consisting lead best result. addition, usage segmentation slightly improves