BUSEM at SemEval-2017 Task 4A Sentiment Analysis with Word Embedding and Long Short Term Memory RNN Approaches

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 777–783, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics BUSEM at SemEval-2017 Task 4 Sentiment Analysis with Word Embedding and Long Short Term Memory RNN Approaches Deger Ayata1, Murat Saraclar 1, Arzucan Ozgur2 1Electrical & Electronical Engineering Department, Bogaziçi University 2Computer Engineering Department, Bogaziçi University Istanbul , Turkey {deger.ayata, murat.saraclar, arzucan.ozgur}@boun.edu.tr Abstract 1 Introduction Sentiment analysis is extracting subjective information from source materials, via natural language processing, computational linguistics, text mining and machine learning. Classification of users’ reviews about a concept or political view may bring different opportunities including customer satisfaction rating, making right recommendations to right target, categorization of users etc. Sentiment Analysis is often referred to as subjectivity analysis, opinion mining and appraisal extraction with some connections to affective computing. Sometimes whole documents are studied as a sentiment unit (Turney and Littman, 2003), but it’s generally agreed that sentiment resides in smaller linguistic units (Pang and Lee, 2008). This paper describes our approach for SemEval-2017 Task 4: Sentiment Analysis in Twitter. We have participated in Subtask A: Message Polarity Classification subtask. We have developed two systems. The first system uses word embeddings for feature representation and Support Vector Machine (SVM), Random Forest (RF) and Naive Bayes (NB) algorithms for classification Twitter messages into negative, neutral and positive polarity. The second system is based on Long Short Term Memory Recurrent Neural Networks (LSTM) and uses word indexes as sequence of inputs for feature representation. The remainder of this article is structured as follows: Section 2 contains information about the system description and Section 3 explains methods, models, tools and software packages used in this work. Test cases and datasets are explained in Section 4. Results are given in Section 5 with discussions. Finally, section 6 summarizes the conclusions and future work. 2 System Description We have developed two independent systems. The first system is word embedding centric and described in subsection 2.1. The second is LSTM based and described in subsection 2.2. Further details about both systems are given in Section 3. 2.1 Word Embedding based System Description In word embedding centric system approach, each word in a tweet is represented with a vector. Tweets consist of words and vectorial values of words (word vectors) are used to represent tweets as vectorial values. Word Embedding system framework is shown Figure 1. Two methods are used to obtain word vectors in this work. The first method is based on generating word vectors via constructing a word2vec model from semeval corpus as depicted in Figure 1 steps 1 and 2. The second This paper describes our approach for SemEval-2017 Task 4: Sentiment Analysis in Twitter. We have participated in Subtask A: Message Polarity Classification subtask and developed two systems. The first system uses word embeddings for feature representation and Support Vector Machine, Random Forest and Naive Bayes algorithms for the classification of Twitter messages into negative, neutral and positive polarity. The second system is based on …

Content truncated to 500 words