Negative Statements Considered Useful

  • 2020-01-13 17:49:37
  • Hiba Arnaout, Simon Razniewski, Gerhard Weikum
  • 9

Abstract

Knowledge bases (KBs), pragmatic collections of knowledge about notableentities, are an important asset in applications such as search, questionanswering and dialogue. Rooted in a long tradition in knowledge representation,all popular KBs only store positive information, while they abstain from takingany stance towards statements not contained in them. In this paper, we make the case for explicitly stating interesting statementswhich are not true. Negative statements would be important to overcome currentlimitations of question answering, yet due to their potential abundance, anyeffort towards compiling them needs a tight coupling with ranking. We introducetwo approaches towards compiling negative statements. (i) In peer-basedstatistical inferences, we compare entities with highly related entities inorder to derive potential negative statements, which we then rank usingsupervised and unsupervised features. (ii) In query-log-based text extraction,we use a pattern-based approach for harvesting search engine query logs.Experimental results show that both approaches hold promising and complementarypotential. Along with this paper, we publish the first datasets on interestingnegative information, containing over 1.1M statements for 100K popular Wikidataentities.

 

Quick Read (beta)

loading the full paper ...