On the Cost of Negation for Dynamic Pruning


Negated query terms allow documents containing such terms to be filtered out of a search results list, supporting disambiguation. In this work, the effect of negation on the efficiency of disjunctive, top-$k$ retrieval is examined. First, we show how negation can be integrated efficiently into two popular dynamic pruning algorithms. Then, we explore the efficiency of our approach, and show that while often efficient, negation can negatively impact the dynamic pruning effectiveness for certain queries.

Proceedings of the 40th European Conference on Information Retrieval (ECIR 2018)