A New Framework for Personal Name Disambiguation

Abstract

In this paper, we study the problem of personal name disambiguation (NED). We develop a framework to address the three challenges in personal name disambiguation: (1) identification of referential ambiguity; (2) identification of lexical ambiguity; and (3) predicting the NIL value, that is the value when a named entity cannot be mapped to a knowledge base. Our framework includes extractor, searcher and disambiguator. Experimental results evaluated on real-world data sets show that our framework and algorithm provide accuracy in personal name linking up to 92%, which is higher than the accuracy of previously developed algorithms.