FBK: Exploiting Phrasal and Contextual Clues for Negation Scope Detection

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 340–346, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics FBK: Exploiting Phrasal and Contextual Clues for Negation Scope Detection Md. Faisal Mahbub Chowdhury † ‡ ‡ Fondazione Bruno Kessler (FBK-irst), Trento, Italy † University of Trento, Italy chowdhury@fbk.eu Abstract Automatic detection of negation cues along with their scope and corresponding negated events is an important task that could bene- fit other natural language processing (NLP) tasks such as extraction of factual information from text, sentiment analysis, etc. This paper presents a system for this task that exploits phrasal and contextual clues apart from vari- ous token specific features. The system was developed for the participation in the Task 1 (closed track) of the *SEM 2012 Shared Task (Resolving the Scope and Focus of Negation), where it is ranked 3rd among the participating teams while attaining the highest F1 score for negation cue detection. 1 Introduction Negation is a linguistic phenomenon that can al- ter the meaning of a textual segment. While auto- matic detection of negation expressions (i.e. cues) in free text has been a subject of research interest for quite some time (e.g. Chapman et al. (2001), Elkin et al. (2005) etc), automatic detection of full scope of negation is a relatively new topic (Morante and Daelemans, 2009; Councill et al., 2010). Detec- tion of negation cues, their scope and corresponding negated events in free text could improve accuracy in other natural language processing (NLP) tasks such as extraction of factual information from text, senti- ment analysis, etc (Jia et al., 2009; Councill et al., 2010). In this paper, we present a system that was de- veloped for the participation in the Scope Detection task of the *SEM 2012 Shared Task1. The proposed system exploits phrasal and contextual clues apart from various token specific features. Exploitation of phrasal clues is not new for negation scope de- tection. But the way we encode this information (i.e. the features for phrasal clues) is novel and dif- fers completely from the previous work (Councill et al., 2010; Morante and Daelemans, 2009). More- over, the total number of features that we use is also comparatively lower. Furthermore, to the best of our knowledge, automatic negated event/property iden- tification has not been explored prior to the *SEM 2012 Shared Task. So, our proposed approach for this particular sub-task is another contribution of this paper. The remainder of this paper is organised as fol- lows. First, we describe the scope detection task as well as the accompanying datasets in Section 2. Then in Section 3, we present how we approach the task. Following that, in Section 4, various empiri- cal results and corresponding analyses are discussed. Finally, we summarize our work and discuss how the system can be further improved in Section 5. 2 Task Description: Scope Detection The Scope Detection task (Task 1) of *SEM 2012 Shared Task deals with intra-sentential (i.e. con- text is single sentence) negations. According to the guidelines …

Content truncated to 500 words