DAEDALUS at SemEval-2014 Task 9: Comparing Approaches for Sentiment Analysis in Twitter

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 218–222, Dublin, Ireland, August 23-24, 2014. DAEDALUS at SemEval-2014 Task 9: Comparing Approaches for Sentiment Analysis in Twitter Julio Villena-Román Daedalus, S.A. jvillena@daedalus.es Janine García-Morera Daedalus, S.A. jgarcia@daedalus.es José Carlos González-Cristóbal Universidad Politécnica de Madrid josecarlos.gonzalez@upm.es Abstract This paper describes our participation at SemEval- 2014 sentiment analysis task, in both contextual and message polarity classification. Our idea was to com- pare two different techniques for sentiment analysis. First, a machine learning classifier specifically built for the task using the provided training corpus. On the other hand, a lexicon-based approach using natural language processing techniques, developed for a ge- neric sentiment analysis task with no adaptation to the provided training corpus. Results, though far from the best runs, prove that the generic model is more robust as it achieves a more balanced evaluation for message polarity along the different test sets. 1 Introduction SemEval1 is an international competitive evalua- tion workshop on semantic related tasks. Among the ten different tasks that have been proposed in 2014, Task 9 at SemEval-20142 focuses on sen- timent analysis in Twitter. Sentiment analysis could be described as the application of natural language processing and text analytics to identify and extract subjective information from texts. Given a message in Eng- lish, the objective is to determine if the text ex- presses a positive, negative or neutral sentiment in that context. It is a major technological challenge and the task is so hard that even humans often disagree on the sentiment of a given text, as issues that one individual may find acceptable or relevant may not be the same to others, along with multi- lingual aspects and different cultural factors. This work is licensed under a Creative Commons Attribu- tion 4.0 International Licence. Page numbers and proceed- ings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/4.0/ 1 http://alt.qcri.org/semeval2014/ 2 http://alt.qcri.org/semeval2014/task9/ The task defines two subtasks, where the dif- ference is that whereas the output in subtask B must be the message polarity classification, i.e., the global polarity of the whole message, subtask A is focused on contextual polarity disambigua- tion, i.e., the message contains a marked instance of a word or phrase and the expected output must be the polarity of that specific instance within the whole message. Daedalus (2014) is a leading provider of lan- guage-based solutions in Spain, and long-time participants in different research conferences and evaluation workshops such as CLEF (2014) and NTCIR (2014), in many different tasks including sentiment analysis (Villena-Román et al., 2008; Villena-Román et al., 2012). This paper describes our participation in both contextual (subtask A) and message (subtask B) polarity classification. The main idea behind our participation is to compare two different tech- niques for sentiment analysis: a machine learning approach using the provided corpus to train a model specifically adapted to that scenario against a lexicon-based approach using advanced natural language processing techniques for cap- turing the meaning of the text, developed prior to the …

Content truncated to 500 words