ContextWalk: Embedding Networks with Context Information Extracted from News Articles

Chaoran Chen, Mirco Schönfeld, Jürgen Pfeffer

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

Abstract

Extracting meaningful social networks from a large corpus of news articles is challenging. This paper presents a network embedding technique which uses the content of texts as context information for relationships among persons referenced in texts. The goal of our approach is to automatically extract social networks from news articles and to identify structures in these networks. ContextWalk is an algorithmic approach extending random walk-based techniques by adapting a context-aware path sampling. We demonstrate the functionality and performance of the algorithm on an 8,000 nodes network extracted from a large German news dataset. Comparison with state-of-the-art algorithms shows that using the context to identify similar types of relationships significantly improves the algorithm’s ability to detect fine-grained semantic groups.

Original languageEnglish
Title of host publicationDatabase and Expert Systems Applications - 32nd International Conference, DEXA 2021, Proceedings
EditorsChristine Strauss, Gabriele Kotsis, A Min Tjoa, Ismail Khalil
PublisherSpringer Science and Business Media Deutschland GmbH
Pages100-114
Number of pages15
ISBN (Print)9783030864743
DOIs
StatePublished - 2021
Event32nd International Conference on Database and Expert Systems Applications, DEXA 2021 - Virtual, Online
Duration: 27 Sep 202130 Sep 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12924 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference32nd International Conference on Database and Expert Systems Applications, DEXA 2021
CityVirtual, Online
Period27/09/2130/09/21

Keywords

  • Context information
  • News
  • Representation learning
  • Social networks

Fingerprint

Dive into the research topics of 'ContextWalk: Embedding Networks with Context Information Extracted from News Articles'. Together they form a unique fingerprint.

Cite this