KDE-AFFECT at SemEval-2018 Task 1: Estimation of Affects in Tweet by Using Convolutional Neural Network for n-gram

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 156–161 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics KDE-AFFECT at SemEval-2018 Task 1: Estimation of Affects in Tweet by Using Convolutional Neural Network for n-gram Shinnosuke Himeno and Masaki Aono Department of Computer Science and Engineering Toyohashi University of Technology himeno@kde.cs.tut.ac.jp, aono@tut.jp Abstract This paper describes our approach to SemEval-2018 Task1: Estimation of Af- fects in Tweet for 1a and 2a. Our team KDE-AFFECT employs several methods including one-dimensional Convolutional Neural Network for n-grams, together with word embedding and other preprocessing such as vocabulary unification and Emoji conversions into four emotional words. 1 Introduction With the rapid spread of SNS services (e.g. Twit- ter, Facebook, Instagram), massive user opinions have accumulated on the Internet. Among such opinions, it has been observed that not a few SNS contents naturally entail the affects (including joy, anger, sadness, fear) within themselves. Hence, the need to accurately detect the affects is increas- ing year by year. In SemEval-2018 Task 1: Estimation of Affects in Tweet, we have attempted to extend our hori- zon from positive, neutral, and negative polarity estimations in former SemEval sentiment analy- sis in tweet having been held till 2017, to mul- tiple emotions (joy, anger, sadness, and fear) in terms of regression (Task-1 1a) and classification (Task-1 2a). In doing so, we have adopted a stan- dard one-dimensional Convolutional Neural Net- work (CNN), which is believed to be effective for text polarity estimation, where the kernel window size for 1D convolution is analogous to the con- cept of word n-gram. In addition, as most peo- ple have noticed, a tweet has potentially many Emojis to express emotions. In the following, we first briefly survey related work on tweet sentiment analysis including emotion estimation. Then, we describe our system, followed by showing the re- sults returned from the organizer, and finally con- cluding our paper. 2 Related Work Sentiment analysis of tweets has been studied by many researchers from the standpoint of clas- sifying a tweet into either positive or negative polarity, and classifying it into multiple emo- tions (Giachanou and Crestani, 2016; Silva et al., 2016). A supervised approach to polarity clas- sification of a tweet was proposed by Go et al. (2009). They employed Naive Bayes, Maxi- mum Entropy, Support Vector Machine, and sev- eral other machine learning methods for their su- pervised learning. Bravo-Marque et al. (2013) pre- sented an approach using multiple emotion dic- tionaries, while Saif et al. (2016) employed co- occurrence information of words. Severyn et al. (2015) introduced a deep learning approach. Lu et al. (2013) proposed a deep learning method suited for short texts. In SemEval, since 2014, sen- timent analysis tasks using Twitter have been offi- cially conducted, where a variety of methods have been tested (Hagen et al., 2015; Giorgis et al., 2016; Deriu et al., 2016; Rouvier and Favre, 2016; Xu et al., 2016). In SemEval2017 Rosenthal et al. (2017), Cliche et al. (2017) and …

Content truncated to 500 words