PNEL: Pointer Network based End-To-End Entity Linking over Knowledge Graphs

Question Answering systems are generally modelled as a pipeline consisting of\na sequence of steps. In such a pipeline, Entity Linking (EL) is often the first\nstep. Several EL models first perform span detection and then entity\ndisambiguation. In such models errors from the span detection phase cascade to\nlater steps and result in a drop of overall accuracy. Moreover, lack of gold\nentity spans in training data is a limiting factor for span detector training.\nHence the movement towards end-to-end EL models began where no separate span\ndetection step is involved. In this work we present a novel approach to\nend-to-end EL by applying the popular Pointer Network model, which achieves\ncompetitive performance. We demonstrate this in our evaluation over three\ndatasets on the Wikidata Knowledge Graph.\n