NILC_USP: Aspect Extraction using Semantic Labels

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 433–436, Dublin, Ireland, August 23-24, 2014. NILC USP: Aspect Extraction using Semantic Labels Pedro P. Balage Filho and Thiago A. S. Pardo Interinstitutional Center for Computational Linguistics (NILC) Institute of Mathematical and Computer Sciences, University of S˜ao Paulo S˜ao Carlos - SP, Brazil {balage, taspardo}@icmc.usp.br Abstract This paper details the system NILC USP that participated in the Semeval 2014: As- pect Based Sentiment Analysis task. This system uses a Conditional Random Field (CRF) algorithm for extracting the aspects mentioned in the text. Our work added se- mantic labels into a basic feature set for measuring the efficiency of those for as- pect extraction. We used the semantic roles and the highest verb frame as fea- tures for the machine learning. Overall, our results demonstrated that the system could not improve with the use of this se- mantic information, but its precision was increased. 1 Introduction Sentiment analysis, or opinion mining, has gained lots of attention lately. The importance of this field of study is linked with the grown of informa- tion in the internet and the commercial attention it brought. According to Liu et al. (2010), there are two kinds of information available in the internet: facts and opinions. Facts are objective statements about entities and events in the world. Opinions are sub- jective statements that reflect people’s sentiments or perceptions about the entities and events. Ac- cording to Liu, by that time, there was a lot of at- tention on the processing of facts but little work had been done on the processing of opinions. This work is licensed under a Creative Commons At- tribution 4.0 International Licence. Page numbers and pro- ceedings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/4.0/ Three levels of analysis for sentiment analysis are known (Liu, 2012): document level, sentence level and aspect level. The aspect-based sentiment analysis is the name of the research topic that aims to extract the sentiments about the aspects present in the text. This work presents a system evaluated in the SemEval Task4: Aspect Based Sentiment Analy- sis shared task (Pontiki et al., 2014). Our system participated only in subtask 1: Aspect Term Ex- traction. In this subtask, given a text, the system should extract all aspects that are present. There were two different domains for this task: restau- rants and laptops. The goal of our system was to verify how se- mantic labels used in machine learning classifica- tion would improve the aspect extraction task. For this goal, we used two kinds of semantic labels: the semantic roles (Palmer et al., 2010) and the se- mantic frames (Baker et al., 1998). Liu et al. (2012) categorizes the works for as- pect extraction in four types, regarding the ap- proach they follow, using: frequent terms, infre- quent terms, machine learning, and topic model- ing. This work uses a machine learning approach that consists in training a sequential labeling algo- rithm for aspect detection and extraction. In …

Content truncated to 500 words