Fake News Detection Through Multi-Perspective Speaker Profiles

Yunfei Long, Qin Lu, Rong Xiang, Minglei Li, Chu-ren Huang

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

Abstract

Automatic fake news detection is an im-portant, yet very challenging topic. Tradi-tional methods using lexical features haveonly very limited success.This paperproposes a novel method to incorporatespeaker profiles into an attention basedLSTM model for fake news detection.Speaker profiles contribute to the model intwo ways. One is to include them in the at-tention model. The other includes them asadditional input data. By adding speakerprofiles such as party affiliation, speakertitle, location and credit history, our modeloutperforms the state-of-the-art method by14.5% in accuracy using a benchmark fakenews detection dataset. This proves thatspeaker profiles provide valuable informa-tion to validate the credibility of news articles.
Original languageEnglish
Title of host publicationProceedings of the The 8th International Joint Conference on Natural Language Processing
Place of PublicationTaipei, Taiwan
PublisherAsian Federation of Natural Language Processing
Pages252-256
Number of pages5
Volume2
ISBN (Print)978-1-948087-01-8
Publication statusPublished - 27 Nov 2017
EventThe 8th International Joint Conference on Natural Language Processing - Taipei, Taiwan
Duration: 27 Nov 20171 Dec 2017

Conference

ConferenceThe 8th International Joint Conference on Natural Language Processing
CountryTaiwan
CityTaipei
Period27/11/171/12/17

Cite this