Tree Kernel-based Negation and Speculation Scope Detection with Structured Syntactic Parse Features

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 968–976, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Tree Kernel-based Negation and Speculation Scope Detection with Structured Syntactic Parse Features Bowei Zou Guodong Zhou Qiaoming Zhu* Natural Language Processing Lab, School of Computer Science and Technology Soochow University, Suzhou, 215006, China zoubowei@gmail.com, {gdzhou,qmzhu}@suda.edu.cn Abstract Scope detection is a key task in information ex- traction. This paper proposes a new approach for tree kernel-based scope detection by using the structured syntactic parse information. In addi- tion, we have explored the way of selecting compatible features for different part-of-speech cues. Experiments on the BioScope corpus show that both constituent and dependency structured syntactic parse features have the advantage in capturing the potential relationships between cues and their scopes. Compared with the state of the art scope detection systems, our system achieves substantial improvement.* 1 Introduction The task of scope detection is to detect the linguis- tic scope dominated by a specific cue. Current re- searches in this field focus on two semantic as- pects: negation and speculation. The negative scope detection is to detect the linguistic scope which is repudiated by a negative word (viz., nega- tive cue, e.g., “not”). In other side, the speculative scope detection is to detect the uncertain part in a sentence corresponding to the speculative word (viz., speculative cue, e.g., “seems”). See the sen- tence 1) below, the negative cue “not” dominates the scope of “not expensive”. Similarly, the specu- lative cue “possible” in sentence 2) dominates the uncertain scope “the possible future scenarios”. 1) The chair is [not expensive] but comfortable. 2) Considering all that we have seen, what are now [the possible future scenarios]? *�Corresponding�author� The negative and speculative scope detection task consists of two basic stages. The first one is to identify the sentences involving negative or specu- lative meaning. The second stage is to detect the linguistic scope of the cue in sentences (Velldal et al, 2012). In this paper, we focus on the second stage. That is, by given golden cues, we detect their linguistic scopes. We propose a tree kernel-based negation and speculation scope detection with structured syntac- tic parse features. In detail, we regard the scope detection task as a binary classification issue, which is to classify the tokens in a sentence as be- ing inside or outside the scope. In the basic framework, we focus on the analysis and applica- tion of structured syntactic parse features as fol- lows: Both constituent and dependency syntactic fea- tures have been proved to be effective in scope detection (Özgür et al, 2009; Øvrelid et al, 2010). However, these flat features are hardly to reflect the information implicit in syntactic parse tree structures. Our intuition is that the segments of the syntactic parse tree around a negative or specula- tive cue is effective for scope detection. The relat- ed structures normally underlay the indirect clues to identify the relations between cues and their scopes, e.g., in …

Content truncated to 500 words