Corroborating facts from affirmative statements

Minji Wu, Amélie Marian

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Web sources often provide different and even conflicting information. Simple voting-based strategies have already shown limitations at identifying the correct answer to a user query with the presence of unreliable sources. In order to identify the correct answer, corroboration techniques have been proposed and proved to be effective for such tasks. In this paper, we investigate the corroboration problem in which most or all facts have only affirmative statements from sources. A fact is either true or false, and an affirmative statement from a source indicates its support for a fact being true. Unfortunately, state-of-the-art corroboration techniques rely on conflicting information to differentiate the trustworthiness of the sources and we demonstrate their limitations in our scenario. Different from existing techniques that consider a single trust score for each source, we propose a novel algorithm that utilizes a multi-value trust score toward different subsets of facts. By considering the information entropy of the unknown facts, our algorithm incrementally evaluates facts and updates the estimates on the trust scores for the sources. We conduct experiments using both synthetic and real-world datasets and demonstrate that our algorithm significantly outperforms existing approaches in precision and accuracy.

Original languageEnglish (US)
Title of host publicationAdvances in Database Technology - EDBT 2014
Subtitle of host publication17th International Conference on Extending Database Technology, Proceedings
EditorsVincent Leroy, Vassilis Christophides, Vassilis Christophides, Stratos Idreos, Anastasios Kementsietsidis, Minos Garofalakis, Sihem Amer-Yahia
PublisherOpenProceedings.org, University of Konstanz, University Library
Pages157-168
Number of pages12
ISBN (Electronic)9783893180653
DOIs
StatePublished - 2014
Event17th International Conference on Extending Database Technology, EDBT 2014 - Athens, Greece
Duration: Mar 24 2014Mar 28 2014

Publication series

NameAdvances in Database Technology - EDBT 2014: 17th International Conference on Extending Database Technology, Proceedings

Other

Other17th International Conference on Extending Database Technology, EDBT 2014
Country/TerritoryGreece
CityAthens
Period3/24/143/28/14

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Information Systems
  • Software

Fingerprint

Dive into the research topics of 'Corroborating facts from affirmative statements'. Together they form a unique fingerprint.

Cite this