TeamCEN at SemEval-2018 Task 1: Global Vectors Representation in Emotion Detection

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 334–338 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics TeamCEN at SemEval-2018 Task 1: Global Vectors Representation in Emotion Detection Anon George, Barathi Ganesh HB, Anand Kumar M and Soman KP Center for Computational Engineering and Networking (CEN) Amrita School of Engineering, Coimbatore Amrita Vishwa Vidyapeetham, India anongeorge007@gmail.com, barathiganesh.hb@gmail.com Abstract Emotions are a way of expressing human sen- timents. In the modern era, social media is a platform where we convey our emotions. These emotions can be joy, anger, sadness and fear. Understanding the emotions from the written sentences is an interesting part in knowing about the writer. In the amount of digital language shared through social media, a considerable amount of data reflects the sen- timent or emotion towards some product, per- son and organization. Since these texts are from users with diverse social aspects, these texts can be used to enrich the application re- lated to the business intelligence. More than the sentiment, identification of intensity of the sentiment will enrich the performance of the end application. In this paper we experimented the intensity prediction as a text classification problem that evaluates the distributed repre- sentation text using aggregated sum and di- mensionality reduction of the glove vectors of the words present in the respective texts. 1 Introduction Emotion detection from text has been an impor- tant task in recent years since the development of Natural Language Processing. Finding the affect in tweet enlarges the business intelligence towards the consumer behaviour analysis, peoples likeness towards the person, organization and policies of the government. Intensity of the sentiment present in the text can be predicted by performing a text classifica- tion by taking texts as the observations and their intensity classes or scores as a target labels. Rep- resentation is a key part in any text classification task which can affect the further feature extrac- tions and predictions (Ganesh et al., 2016; So- man et al., 2016; B. et al., 2016). A bad repre- sentation of the word vector can make the further predictions fruitless. Hence we use a meaning- ful representation of words, that is, global vectors (Glove) for the representation of words into vec- tors of a tweet. This paper explains the use of global vectors representation for representing the words in a tweet and their further processing using machine learning techniques for classification and regression tasks. SemEval-2018 Task 1 provided the twitter corpus for classification and regression tasks for the emotion detection. After representing the words in the tweets, the vector for tweets is obtained by computing the ag- gregated sum and dimensionality reduction. The final tweet vectors are used for further classifica- tion and regression. 2 Global Vectors Global Vectors (GloVe) creates word vectors and it is an unsupervised machine learning tech- nique(Pennington et al., 2014). Unlike word2vec representations like skip-gram and continuous bag of words (CBOW), word-word co occurrence stat- ics is taken for the vector representation. It also retains …

Content truncated to 500 words