Detecting Optimism in Tweets using Knowledge Distillation and Linguistic Analysis of Optimism

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 2032–2041 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 2032 Detecting Optimism in Tweets using Knowledge Distillation and Linguistic Analysis of Optimism S, tefan Cobeli∗, Bogdan Iordache†, Shweta Yadav∗, Cornelia Caragea∗, Liviu Dinu†, Dragos, Iliescu† ∗University of Illinois at Chicago, Illinois, US †University of Bucharest, Bucharest, Romania ∗{scobel2, shwetay, cornelia}@uic.edu †{iordache.bogdan1998@gmail.com, liviu.p.dinu@gmail.com, dragos.iliescu@fpse.unibuc.ro} Abstract Finding the polarity of feelings in texts is a far-reaching task. Whilst the field of natural language processing has established sentiment analysis as an alluring problem, many feelings are left uncharted. In this study, we analyze the optimism and pes- simism concepts from Twitter posts to effectively understand the broader dimension of psychological phenomenon. Towards this, we carried a systematic study by first exploring the linguistic peculiarities of optimism and pessimism in user-generated content. Later, we devised a multi-task knowledge distillation framework to simultaneously learn the target task of optimism detection with the help of the auxiliary task of sentiment analysis and hate speech detection. We evaluated the performance of our proposed approach on the benchmark Optimism/Pessimism Twitter dataset. Our extensive experiments show the superior- ity of our approach in correctly differentiating between optimistic and pessimistic users. Our human and automatic evaluation shows that sentiment analysis and hate speech detection are beneficial for optimism/pessimism detection. 1. Introduction The optimism-pessimism continuum is a stable person- ality trait, represented by the dispositional tendency to hold generalized and positive expectancies even when confronted with adversity and stress (Scheier et al., 2001) The optimism-pessimism continuum has been intensively studied because of its links to a number of important outcomes, both in the realm of perfor- mance (e.g., job performance; (Kumar et al., 2017)), and satisfaction (e.g., life satisfaction; (Extremera et al., 2009)). Among all the individual differences, optimism-pessimism is one of the strongest connected to happiness (Alarcon et al., 2013), stress management (Carver et al., 2010) and ultimately mental (Scheier and Carver, 1992) and, perhaps unexpectedly, physical wellbeing (Rasmussen et al., 2009). The mechanisms of this connection are not yet all clear, but we know now that optimism has a versatile role in different con- text: it acts as a strong protective factor against adver- sity (Gallagher et al., 2020), it acts as a distal predictor and encourages effective coping against and recovery from anxiety, depression, traumatic events and suicidal inclinations (Achat et al., 2000; Prati and Pietrantoni, 2009), and it has a strong cross-over effect to close rel- atives and acquaintances (Scheier et al., 2001; Peterson and Bossio, 2001). The fundamental mindset, i.e., the generalized positive or negative expectancy, or "frame" (McKenzie and Nel- son, 2003), held by a person who is characterized by optimism or pessimism influences in a profound man- ner the ways in which that person observes, processes, understands and reacts to the environment. Knowledge of this mindset therefore permits both immediate and long term predictions of human behavior. Assessing dispositional optimism-pessimism on the fly is there- fore …

Content truncated to 500 words