*SEM 2012 Shared Task: Resolving the Scope and Focus of Negation

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 265–274, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics *SEM 2012 Shared Task: Resolving the Scope and Focus of Negation Roser Morante CLiPS - University of Antwerp Prinsstraat 13, B-2000 Antwerp, Belgium Roser.Morante@ua.ac.be Eduardo Blanco Lymba Corporation Richardson, TX 75080 USA eduardo@lymba.com Abstract The Joint Conference on Lexical and Compu- tational Semantics (*SEM) each year hosts a shared task on semantic related topics. In its first edition held in 2012, the shared task was dedicated to resolving the scope and focus of negation. This paper presents the specifica- tions, datasets and evaluation criteria of the task. An overview of participating systems is provided and their results are summarized. 1 Introduction Semantic representation of text has received consid- erable attention these past years. While early shal- low approaches have been proven useful for several natural language processing applications (Wu and Fung, 2009; Surdeanu et al., 2003; Shen and La- pata, 2007), the field is moving towards analyzing and processing complex linguistic phenomena, such as metaphor (Shutova, 2010) or modality and nega- tion (Morante and Sporleder, 2012). The *SEM 2012 Shared Task is devoted to nega- tion, specifically, to resolving its scope and focus. Negation is a grammatical category that comprises devices used to reverse the truth value of proposi- tions. Broadly speaking, scope is the part of the meaning that is negated and focus the part of the scope that is most prominently or explicitly negated (Huddleston and Pullum, 2002). Although negation is a very relevant and complex semantic aspect of language, current proposals to annotate meaning ei- ther dismiss negation or only treat it in a partial man- ner. The interest in automatically processing nega- tion originated in the medical domain (Chapman et al., 2001), since clinical reports and discharge summaries must be reliably interpreted and indexed. The annotation of negation and hedge cues and their scope in the BioScope corpus (Vincze et al., 2008) represented a pioneering effort. This corpus boosted research on scope resolution, especially since it was used in the CoNLL 2010 Shared Task (CoNLL ST 2010) on hedge detection (Farkas et al., 2010). Negation has also been studied in sentiment analy- sis (Wiegand et al., 2010) as a means to determine the polarity of sentiments and opinions. Whereas several scope detectors have been de- veloped using BioScope (Morante and Daelemans, 2009; Velldal et al., 2012), there is a lack of cor- pora and tools to process negation in general domain texts. This is why we have prepared new corpora for scope and focus detection. Scope is annotated in Conan Doyle stories (CD-SCO corpus). For each negation, the cue, its scope and the negated event, if any, are marked as shown in example (1a). Focus is annotated on top of PropBank, which uses the WSJ section of the Penn TreeBank (PB-FOC corpus). Fo- cus annotation is restricted to verbal negations an- notated with MNEG in PropBank, and all the words belonging to a semantic role …

Content truncated to 500 words