Guilt by Association: Emotion Intensities in Lexical Representations

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 9911–9917 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 9911 Guilt by Association: Emotion Intensities in Lexical Representations Shahab Raji Rutgers University shahab.raji@rutgers.edu Gerard de Melo Hasso Plattner Institute/University of Potsdam http://gerard.demelo.org Abstract What do linguistic models reveal about the emotions associated with words? In this study, we consider the task of estimating word-level emotion intensity scores for specific emotions, exploring unsupervised, supervised, and fi- nally a self-supervised method of extracting emotional associations from pretrained vectors and models. Overall, we find that linguistic models carry substantial potential for inducing fine-grained emotion intensity scores, showing a far higher correlation with human ground truth ratings than state-of-the-art emotion lexi- cons based on labeled data. 1 Introduction There has been substantial research on methods to label words with associated emotions. Crowd- sourcing approaches have been used to compile databases to study the nexus between them (Mo- hammad and Turney, 2013; Mohammad, 2018; Shoeb and de Melo, 2020). Another strategy, adopted by the well-known DepecheMood (Staiano and Guerini, 2014) and DepecheMood++ (Araque et al., 2018) lexicons is to apply simple statistical methods to emotionally tagged data crawled from specific online sources. In this paper, we consider the question: What do linguistic models reveal about the emotions asso- ciated with words? Word vectors (Mikolov et al., 2013; Pennington et al., 2014) have often been eval- uated on standard word relatedness benchmarks. In this paper, we instead explore to what extent they encode emotional associations. Earlier methods (Strapparava and Mihalcea, 2008; Mac Kim et al., 2010) used corpus statis- tics in tandem with dimensionality reduction tech- niques for emotion prediction. Shoeb et al. (2019) considered word embeddings trained on emojis for emotion analysis. Rothe et al. (2016) proposed predicting sentiment polarity ratings from word vectors, while other studies (Buechel and Hahn, 2018b; Buechel et al., 2018) predicted valence, arousal, and dominance using supervised deep neu- ral networks. Khosla et al. (2018) proposed incor- porating valence, arousal, and dominance ratings as additional signals into dense word embeddings. Buechel and Hahn (2018a) showed that emotion intensity scores can be predicted based on a lexicon providing valence–arousal–dominance ratings. The task of emotional association of words has been studied on other languages as well. Sidorov et al. (2012) presents a dataset of Spanish words labelled with Ekman’s six emotions, while others explored cross-lingual propagation from one language to an- other (Abdaoui et al., 2017) or to hundreds of other languages (Ramachandran and de Melo, 2020). We show that we can make use of linguistic models to obtain high correlations with emotion intensities without any need for manual ratings. While this work focuses on English, the methodology can be applied to data in any language. Overview. We introduce the task and experimen- tal setup in Section 2. Then, section 3 presents our techniques to address this task based on linguistic models. We provide detailed empirical evaluation results in Section 4. Finally, Section 5 concludes the paper by discussing …

Content truncated to 500 words