Entity Disambiguation with Entity Definitions
This addresses a critical limitation in entity disambiguation for NLP applications, though it is incremental by building on existing generative and extractive methods.
The paper tackled the problem of entity disambiguation by using more expressive textual representations instead of just Wikipedia titles, achieving new state-of-the-art results on 2 out of 6 benchmarks and improving generalization over unseen patterns.
Local models have recently attained astounding performances in Entity Disambiguation (ED), with generative and extractive formulations being the most promising research directions. However, previous works limited their studies to using, as the textual representation of each candidate, only its Wikipedia title. Although certainly effective, this strategy presents a few critical issues, especially when titles are not sufficiently informative or distinguishable from one another. In this paper, we address this limitation and investigate to what extent more expressive textual representations can mitigate it. We thoroughly evaluate our approach against standard benchmarks in ED and find extractive formulations to be particularly well-suited to these representations: we report a new state of the art on 2 out of 6 benchmarks we consider and strongly improve the generalization capability over unseen patterns. We release our code, data and model checkpoints at https://github.com/SapienzaNLP/extend.