Detecting hedges scope based on phrase structures and dependency structures

Abstract

To distinguish facts from unreliable or uncertain information, hedges have to be identified. This paper presents an approach to hedges scope detection based on phrase structures and dependency structures. First, phrase structures and dependency structures are used for hedges scope detection respectively. Phrase structures are adapted as important features for hedges scope detection by a machining learning method. Dependency structures are used to detect hedges scope by a rule-based method. Then, the phrase-based system and the dependency-based system are combined by a Conditional Random Field (CRF)-based model, which simply extends the feature vectors with the scope tags generated by the two individual phrase-based and dependency-based systems. Experiments on the CoNLL-2010 biological corpus show that our model achieves F-scores of 55.47% on hedges scope detection based on phrase structures using machine learning and 55.67% based on dependency structures using manual rules, and 58.97% based on dependency structures and phrase structures using our combined method. The analysis results show that phrase structures and dependency structures are both effective for hedges scope detection and their combination can improve the scope detection performance further.