Geometry and Analogies: A Study and Propagation Method for Word Representations

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this paper we discuss the well-known claim that language analogies yield almost parallel vector differences in word embeddings. On the one hand, we show that this property, while it does hold for a handful of cases, fails to hold in general especially in high dimension, using the best known publicly available word embeddings. On the other hand, we show that this property is not crucial for basic natural language processing tasks such as text classification. We achieve this by a simple algorithm which yields updated word embeddings where this property holds: we show that in these word representations, text classification tasks have about the same performance.

Original languageEnglish
Title of host publicationStatistical Language and Speech Processing - 7th International Conference, SLSP 2019, Proceedings
EditorsCarlos Martín-Vide, Matthew Purver, Senja Pollak
PublisherSpringer Science and Business Media Deutschland GmbH
Pages100-111
Number of pages12
ISBN (Print)9783030313715
DOIs
Publication statusPublished - 1 Jan 2019
Event7th International Conference on Statistical Language and Speech Processing, SLSP 2019 - Ljubljana, Slovenia
Duration: 14 Oct 201916 Oct 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11816 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference7th International Conference on Statistical Language and Speech Processing, SLSP 2019
Country/TerritorySlovenia
CityLjubljana
Period14/10/1916/10/19

Fingerprint

Dive into the research topics of 'Geometry and Analogies: A Study and Propagation Method for Word Representations'. Together they form a unique fingerprint.

Cite this