UABCoRAL: A Preliminary study for Resolving the Scope of Negation

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 275–281, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics UABCoRAL: A Preliminary study for Resolving the Scope of Negation Binod Gyawali, Thamar Solorio CoRAL Lab Department of Computer and Information Sciences University of Alabama at Birmingham Birmingham, Alabama, USA {bgyawali,solorio}@cis.uab.edu Abstract This paper describes our participation in the closed track of the *SEM 2012 Shared Task of finding the scope of negation. To perform the task, we propose a system that has three components: negation cue detection, scope of negation detection, and negated event detec- tion. In the first phase, the system creates a lexicon of negation signals from the training data and uses the lexicon to identify the nega- tion cues. Then, it applies machine learning approaches to detect the scope and negated event for each negation cue identified in the first phase. Using a preliminary approach, our system achieves a reasonably good accuracy in identifying the scope of negation. 1 Introduction All human language samples, either written or spo- ken, contain some information in negated form. In tasks such as information retrieval, sometimes, we should consider only the positive information of an event and disregard its negation information, and vice versa. For example, while searching for the pa- tients with diabetes, we should not include a patient who has a clinical report saying No symptoms of di- abetes were observed. Thus, finding the negation and its scope is important in tasks where the nega- tion and assertion information need to be treated dif- ferently. However, most of the systems developed for processing natural language data do not consider negations present in the sentences. Although various works (Morante et al., 2008; Morante and Daele- mans, 2009; Li et al., 2010; Councill et al., 2010; Apostolova et al., 2011) have dealt with the identifi- cation of negations and their scope in sentences, this is still a challenging task. The first task in *SEM 2012 Shared Task (Morante and Blanco, 2012) is concerned with finding the scope of negation. The task includes identifying: i) negation cues, ii) scope of negation, and iii) negated event for each negation present in the sentences. Negation cue is a word, part of a word, or a combination of words that carries the negation information. Scope of negation in a sentence is the longest group of words in the sentence that is influenced by the negation cue. Negated event is the shortest group of words that is actually affected by the negation cue. In Example (1) below, word no is a negation cue, the discontinuous word sequences ‘I gave him’ and ‘sign of my occupation’ are the scopes, and ‘gave’ is the negated event. (1) I [gave] him no sign of my occupation. In this paper, we propose a system to detect the scope of negation for the closed track of *SEM 2012 Shared Task. Our system uses a combination of a rule based approach, and a machine learning ap- proach. We use a …

Content truncated to 500 words