GWU NLP Lab at SemEval-2019 Task 3 : EmoContext: Effectiveness ofContextual Information in Models for Emotion Detection inSentence-level at Multi-genre Corpus

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 230–235 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 230 GWU NLP Lab at SemEval-2019 Task 3 : EmoContext: Effective Contextual Information in Models for Emotion Detection in Sentence-level in a Multigenre Corpus Shabnam Tafreshi George Washington University Department of Computer Science Washington, DC shabnamt@gwu.edu Mona Diab AWS AI George Washington University Department of Computer Science Washington, DC mtdiab@gwu.edu Abstract In this paper we present an emotion classifier models that submitted to the SemEval-2019 Task 3 : EmoContext. The task objective is to classify emotion (i.e. happy, sad, angry) in a 3-turn conversational data set. We formu- late the task as a classification problem and introduce a Gated Recurrent Neural Network (GRU) model with attention layer, which is bootstrapped with contextual information and trained with a multigenre corpus. We utilize different word embeddings to empirically se- lect the most suited one to represent our fea- tures. We train the model with a multigenre emotion corpus to leverage using all avail- able training sets to bootstrap the results. We achieved overall %56.05 f1-score and placed 144. 1 Introduction In recent studies, deep learning models have achieved top performances in emotion detection and classification. Access to large amount of data has contributed to these high results. Numerous efforts have been dedicated to build emotion clas- sification models, and successful results have been reported. In this work, we combine several popu- lar emotional data sets in different genres, plus the one given for this task to train the emotion model we developed. We introduce a multigenre train- ing mechanism, our intuition to combine differ- ent genres are a) to augment more training data, b) to generalize detection of emotion. We uti- lize Portable textual information such as subjec- tivity, sentiment, and presence of emotion words, because emotional sentences are subjective and af- fectual states like sentiment are strong indicator for presence of emotion. The rest of this paper is structured as followings: section 2 introduce our neural net model, in sec- tion 3 we explain the experimental setup and data that is been used for training and development sets, section 4 discuss the results and analyze the er- rors, section 5 describe related works, section 6 conclude our study and discuss future direction. 2 Model Description Gates Recurrent Neural Network (GRU) (Cho et al., 2014; Chung et al., 2015) and attention layer are used in sequential NLP problems and successful results are reported in different studies. Figure 1 shows the diagram of our model. 1 GRU- has been widely used in the literature to model sequential problems. RNN applies the same set of weights recursively as follow: ht = f(Wxt + Uht−1 + b) (1) GRU is very similar to LSTM with the following equations: rt = σ(W r xt + U rht−1 + br) (2) zt = σ(W z xt + U zht−1 + bz) (3) ˆht = tanh(Wxt + rt × U ˆhht−1 + b ˆh) (4) …

Content truncated to 500 words