Automatic identification of best answers in online enquiry communities

  • Authors:
  • Grégoire Burel;Yulan He;Harith Alani

  • Affiliations:
  • Knowledge Media Institute, The Open University, Milton Keynes, UK;Knowledge Media Institute, The Open University, Milton Keynes, UK;Knowledge Media Institute, The Open University, Milton Keynes, UK

  • Venue:
  • ESWC'12 Proceedings of the 9th international conference on The Semantic Web: research and applications
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Online communities are prime sources of information. The Web is rich with forums and Question Answering (Q&A) communities where people go to seek answers to all kinds of questions. Most systems employ manual answer-rating procedures to encourage people to provide quality answers and to help users locate the best answers in a given thread. However, in the datasets we collected from three online communities, we found that half their threads lacked best answer markings. This stresses the need for methods to assess the quality of available answers to: 1) provide automated ratings to fill in for, or support, manually assigned ones, and; 2) to assist users when browsing such answers by filtering in potential best answers. In this paper, we collected data from three online communities and converted it to RDF based on the SIOC ontology. We then explored an approach for predicting best answers using a combination of content, user, and thread features. We show how the influence of such features on predicting best answers differs across communities. Further we demonstrate how certain features unique to some of our community systems can boost predictability of best answers.