CHOLAN: A Modular Approach for Neural Entity Linking on Wikipedia and Wikidata

In this paper, we propose CHOLAN, a modular approach to target end-to-end\nentity linking (EL) over knowledge bases. CHOLAN consists of a pipeline of two\ntransformer-based models integrated sequentially to accomplish the EL task. The\nfirst transformer model identifies surface forms (entity mentions) in a given\ntext. For each mention, a second transformer model is employed to classify the\ntarget entity among a predefined candidates list. The latter transformer is fed\nby an enriched context captured from the sentence (i.e. local context), and\nentity description gained from Wikipedia. Such external contexts have not been\nused in the state of the art EL approaches. Our empirical study was conducted\non two well-known knowledge bases (i.e., Wikidata and Wikipedia). The empirical\nresults suggest that CHOLAN outperforms state-of-the-art approaches on standard\ndatasets such as CoNLL-AIDA, MSNBC, AQUAINT, ACE2004, and T-REx.\n