Output details

Speculation and negation: Rules, rankers, and the role of syntax

<22> This article, published in the foremost journal in natural language processing, presents techniques for analysing speculation and negation. Proper treatment of such aspects is essential in order to prevent hypothesised or false propositions entering a knowledge base (for example, when performing information extraction over biomedical literature). Previous approaches treat the task as sequence-labelling using surface level features. Instead we examine the utility of syntactic representations through dependency structures and constituent trees. The results published in this article demonstrate that our combination of heuristics with an automatically-learned ranking function yields the strongest performance on benchmark datasets.