Entity typing using distributional semantics and DBpedia

Marieke van Erp*, Piek Vossen

*Corresponding author for this work

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

14 Downloads (Pure)

Abstract

Recognising entities in a text and linking them to an external resource is a vital step in creating a structured resource (e.g. a knowledge base) from text. This allows semantic querying over a dataset, for example selecting all politicians or football players. However, traditional named entity recognition systems only distinguish a limited number of entity types (such as Person, Organisation and Location) and entity linking has the limitation that often not all entities found in a text can be linked to a knowledge base. This creates a gap in coverage between what is in the text and what can be annotated with fine grained types. This paper presents an approach to detect entity types using DBpedia type information and distributional semantics. The distributional semantics paradigm assumes that similar words occur in similar contexts. We exploit this by comparing entities with an unknown type to entities for which the type is known and assign the type of the most similar set of entities to the entity with the unknown type. We demonstrate our approach on seven different named entity linking datasets. To the best of our knowledge, our approach is the first to combine word embeddings with external type information for this task. Our results show that this task is challenging but not impossible and performance improves when narrowing the search space by adding more context to the entities in the form of topic information.

Original languageEnglish
Title of host publicationKnowledge Graphs and Language Technology
Subtitle of host publicationISWC 2016 International Workshops: KEKI and NLP&DBpedia, Kobe, Japan, October 17-21, 2016, Revised Selected Papers
EditorsHeiko Paulheim, Sebastian Hellmann, John P. McCrae, Christian Chiarcos, Pablo Mendes, Hideaki Takeda, Key-Sun Choi, Jorge Gracia, Yoshihiko Hayashi, Seiji Koide, Marieke van Erp
PublisherSpringer Verlag,
Pages102-118
Number of pages17
ISBN (Electronic)9783319687230
ISBN (Print)9783319687223
DOIs
Publication statusPublished - 2017
Event15th International Semantic Web Conference, ISWC 2016 held in conjuction with the 1st Workshop on Knowledge Extraction and Knowledge Integration, KEKI 2016 and 4th NLP and DBpedia Workshop, NLP-DBpedia 2016 - Kobe, Japan
Duration: 17 Oct 201621 Oct 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10579 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference15th International Semantic Web Conference, ISWC 2016 held in conjuction with the 1st Workshop on Knowledge Extraction and Knowledge Integration, KEKI 2016 and 4th NLP and DBpedia Workshop, NLP-DBpedia 2016
Country/TerritoryJapan
CityKobe
Period17/10/1621/10/16

Fingerprint

Dive into the research topics of 'Entity typing using distributional semantics and DBpedia'. Together they form a unique fingerprint.

Cite this