References & Citations
Computer Science > Computation and Language
Title: Specialising Word Vectors for Lexical Entailment
(Submitted on 17 Oct 2017 (v1), last revised 19 Apr 2018 (this version, v2))
Abstract: We present LEAR (Lexical Entailment Attract-Repel), a novel post-processing method that transforms any input word vector space to emphasise the asymmetric relation of lexical entailment (LE), also known as the IS-A or hyponymy-hypernymy relation. By injecting external linguistic constraints (e.g., WordNet links) into the initial vector space, the LE specialisation procedure brings true hyponymy-hypernymy pairs closer together in the transformed Euclidean space. The proposed asymmetric distance measure adjusts the norms of word vectors to reflect the actual WordNet-style hierarchy of concepts. Simultaneously, a joint objective enforces semantic similarity using the symmetric cosine distance, yielding a vector space specialised for both lexical relations at once. LEAR specialisation achieves state-of-the-art performance in the tasks of hypernymy directionality, hypernymy detection, and graded lexical entailment, demonstrating the effectiveness and robustness of the proposed asymmetric specialisation model.
Submission history
From: Ivan Vulić [view email][v1] Tue, 17 Oct 2017 16:39:09 GMT (374kb,D)
[v2] Thu, 19 Apr 2018 13:02:39 GMT (371kb,D)
Link back to: arXiv, form interface, contact.