End-to-End Neural Entity Linking
This addresses the challenge of semantic text understanding and information extraction for researchers and practitioners in NLP, offering a novel approach to improve EL accuracy, though it may be incremental in its neural adaptation of existing concepts.
The paper tackles the problem of Entity Linking (EL) by proposing the first neural end-to-end system that jointly discovers and links entities, eliminating the need for separate Mention Detection and Entity Disambiguation stages. The result shows that this method significantly outperforms popular systems on the Gerbil platform when training data is sufficient, achieving top or competitive accuracy in various scenarios.
Entity Linking (EL) is an essential task for semantic text understanding and information extraction. Popular methods separately address the Mention Detection (MD) and Entity Disambiguation (ED) stages of EL, without leveraging their mutual dependency. We here propose the first neural end-to-end EL system that jointly discovers and links entities in a text document. The main idea is to consider all possible spans as potential mentions and learn contextual similarity scores over their entity candidates that are useful for both MD and ED decisions. Key components are context-aware mention embeddings, entity embeddings and a probabilistic mention - entity map, without demanding other engineered features. Empirically, we show that our end-to-end method significantly outperforms popular systems on the Gerbil platform when enough training data is available. Conversely, if testing datasets follow different annotation conventions compared to the training set (e.g. queries/ tweets vs news documents), our ED model coupled with a traditional NER system offers the best or second best EL accuracy.