UIUC at SemEval-2018 Task 1: Recognizing Affect with Ensemble Models

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 377–384 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics UIUC at SemEval-2018 Task 1: Recognizing Affect with Ensemble Models Abhishek Narwekar Department of Computer Science University of Illinois, Urbana Champaign USA 61820 abhisheknkar@gmail.com Roxana Girju Linguistics Department, Computer Science Department, Beckman Research Institute, University of Illinois, Urbana Champaign USA 61820 girju@illinois.edu Abstract Our submission to the SemEval-2018 Task1: Affect in Tweets shared task competition is a supervised learning model relying on standard lexicon features coupled with word embed- ding features. We used an ensemble of di- verse models, including random forests, gra- dient boosted trees, and linear models, cor- rected for training-development set mismatch. We submitted the system’s output for sub- tasks 1 (emotion intensity prediction), 2 (emo- tion ordinal classification), 3 (valence inten- sity regression) and 4 (valence ordinal clas- sification), for English tweets. We placed 25th, 19th, 24th and 15th in the four subtasks respectively. The baseline considered was an SVM (Support Vector Machines) model with linear kernel on the lexicon and embedding based features. Our system’s final perfor- mance measured in Pearson correlation scores outperformed the baseline by a margin of 2.2% to 14.6% across all tasks. 1 Introduction Affective computing deals with the recognition, interpretation, processing, and simulation of hu- man affects. It is a highly interdisciplinary field at the heart of a broad range of technological ap- plications in health care, media & advertisement, automotive, and others. Although emotions are a fundamental feature of human experience, they have been long ignored by technology development mainly due to their complex and subjective nature, as well as the lack of learning capabilities to detect them. Current affective computing systems focus mainly on fa- cial expressions, body language, speech (tone of voice, rhythm, etc.), keystroke as well as physio- logical input (e.g., heart rate, body temperature) to capture and process changes in a user’s emotional state. However, in environments such as social media and Internet forums, most often the only signal is written language. And since language per se is the smallest portion of human commu- nication (Mehrabian, 1981), emotions are not easy to detect. Although emotion detection is directly related to the more popular task of sentiment analysis, they differ in many respects. Sentiment Analy- sis aims to detect the positive, neutral, or negative orientation of the text, while emotion detection focuses on recognizing and classifying text snip- pets into a set of predefined, more or less univer- sal emotions. Various such classification models have been proposed, two famous ones being Ek- man’s (Ekman, 1997) six basic emotions (anger, happiness, surprise, disgust, sadness, and fear) and Plutchik’s wheel of eight emotions (Plutchik, 2001), where each primary emotion has a polar op- posite (joy, trust, fear, surprise, sadness, anticipa- tion, anger, and disgust). To date, there are many freely available tools for sentiment polarity classification of input text, yet not so many exist for emotion detection. Ma- jor challenges are: (1) the difficulty in …

Content truncated to 500 words