ECNU: Multi-level Sentiment Analysis on Twitter Using Traditional Linguistic Features and Word Embedding Features

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 561–567, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics ECNU: Multi-level Sentiment Analysis on Twitter Using Traditional Linguistic Features and Word Embedding Features Zhihua Zhang, Guoshun Wu, Man Lan∗ Shanghai Key Laboratory of Multidimensional Information Processing Department of Computer Science and Technology, East China Normal University Shanghai 200241, P. R. China {51131201039, 51141201064}@ecnu.cn, mlan@cs.ecnu.edu.cn∗ Abstract This paper reports our submission to task 10 (Sentiment Analysis on Tweet, SAT) (Rosen- thal et al., 2015) in SemEval 2015 , which contains five subtasks, i.e., contextual polar- ity disambiguation (subtask A: expression- level), message polarity classification (subtask B: message-level), topic-based message polar- ity classification and detecting trends towards a topic (subtask C and D: topic-level), and de- termining sentiment strength of twitter terms (subtask E: term-level). For the first four sub- tasks, we built supervised models using tradi- tional features and word embedding features to perform sentiment polarity classification. For subtask E, we first expanded the training data with the aid of external sentiment lexi- cons and then built a regression model to esti- mate the sentiment strength. Despite the sim- plicity of features, our systems rank above the average. 1 Introduction In the past few years, hundreds of millions of peo- ple shared and expressed their opinions through mi- croblogging websites, such as Twitter. The study on this platform is increasingly drawing attention of many researchers and organizations. Given the character limitations on tweets, the sentiment orien- tation classification on tweets is usually analogous to the sentence-level sentiment analysis (Kouloump- is et al., 2011; Kim and Hovy, 2004; Yu and Hatzi- vassiloglou, 2003). However, considering opinions adhering on different topics and expressed by vari- ous expression words in tweets, (Wang et al., 2011; Jiang et al., 2011; Chen et al., 2012) have inves- tigated various ways to settle these target depen- dent issues. Recently, inspired by (Mikolov et al., 2013a) using neural network to construct distributed word representation (word embedding), several re- searchers employed neural network to perform senti- ment analysis. For example, (Kim, 2014; dos Santos and Gatti, 2014) adopted convolutional neural net- works to learn sentiment-bearing sentence vectors, and (Mikolov et al., 2013b) proposed Paragraph vector which outperformed bag-of-words model for sentiment analysis. The task of Sentiment Analysis in Twitter (SAT) in SemEval 2015 consists of five subtasks. The first three subtasks focus on determining the polarity of the given tweet, phrase or topic (i.e., subtask A aims at classifying the sentiment of a marked instance in a given message, subtask B is to determine the po- larity of the whole message and subtask C focuses on identifying the sentiment of the message towards the given topic). The fourth subtask D is to detect the sentiment trends of a given set of messages to- wards a topic from the same period of time. The last subtask E is to predict a score between 0 and 1, which is indicative of the strength of association of twitter …

Content truncated to 500 words