Exploring The Use Of Hybrid Similarity Measure For Author Name Disambiguation

Abstract Name disambiguation has become one of the hard to crack problem in a virtual setup. With each passing day more and more entities with identical features are emerging online making it quite difficult to distinguish them. Digital libraries face similar problems in differentiating publications of similar looking authors. This leads to incorrect attribution of publications thus making the entire effort of indexing publications of individual authors ineffective. This paper proposes a two stage hybrid similarity computation mechanism that combines the best of both the worlds. The proposed method use a token-based similarity score in this first stage of comparison and based on the results of the first stage it uses a character-based similarity score in the second stage. Experimental results obtained on standard datasets indicate that the proposed technique shows a lot of improvements over the existing methods.