Entity-Aspect Linking

Abstract

e availability of entity linking technologies provides a novel way to organize, categorize, and analyze large textual collections in digital libraries. However, in many situations a link to an entity o ers only relatively coarse-grained semantic information. is is problematic especially when the entity is related to several different events, topics, roles, and -more generally -when it has di erent aspects. In this work, we introduce and address the task of entity-aspect linking: given a mention of an entity in a contextual passage, we re ne the entity link with respect to the aspect of the entity it refers to. We show that a combination of di erent features and aspect representations in a learning-to-rank se ing correctly predicts the entity-aspect in 70% of the cases. Additionally, we demonstrate signi cant and consistent improvements using entityaspect linking on three entity prediction and categorization tasks relevant for the digital library community.

Dynamic Named Entity Recognition

“…These progresses have pushed the community to design more challenging NER tasks such as Entity-Aspect Linking, Event Extraction [25] or take over Semantic Role Labeling [20]. These tasks require to exploit the context either by establishing a link to a knowledge base or bound tokens between them with links having a special semantic.…”

“…winning/loosing teams in a basketball match). Related tasks encompass Entity-Aspect Linking (EAL) task [25] in which labels are constrained by fine-grained concepts in a knowledge base, Semantic Role Labeling (SRL) [20] in which labels correspond to roles payed by tokens in their context (verb, subject...) or Event Extraction (EE) [45] in which a token describing an event (called a trigger) is extracted and linked to associated parameters (defined by the 5W1H Who, What, Whom, When, Where and How). Our scenario is different as the class is not specifically constrained by a knowledge base (as in EAL) or need not to be associated to explicit mentions in the input text (as in SRL or EE).…”