EmpathBERT: A BERT-based Framework for Demographic-aware Empathy Prediction

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 3072–3079 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 3072 EMPATHBERT: A BERT-based Framework for Demographic-aware Empathy Prediction Bhanu Prakash Reddy Guda Adobe Research guda@adobe.com Aparna Garimella Adobe Research garimell@adobe.com Niyati Chhaya Adobe Research nchhaya@adobe.com Abstract Affect preferences vary with user demograph- ics, and tapping into demographic information provides important cues about the users’ lan- guage preferences. In this paper, we utilize the user demographics, and propose EMPATH- BERT, a demographic-aware framework for empathy prediction based on BERT. Through several comparative experiments, we show that EMPATHBERT surpasses traditional ma- chine learning and deep learning models, and illustrate the importance of user demograph- ics to predict empathy and distress in user re- sponses to stimulative news articles. We also highlight the importance of affect information in the responses by developing affect-aware models to predict user demographic attributes. 1 Introduction Modeling complex human reactions and affect from text has been a challenging research area with innovations focusing on sentiment and emotion un- derstanding (Picard, 1997; Li and Liu, 2015; Rosen- thal et al., 2017; Socher et al., 2011, 2013). The study of non-trivial human reactions has been lim- ited. These methods, often rooted in psychological theories, have turned out to be more complex in terms of annotation and modeling (Strapparava and Mihalcea, 2007). A critical affective phenomena, empathy, has received surprisingly less attention. Empathy assesses feelings of sympathy towards others, and Distress measures anxiety and discom- fort oriented towards self (Davis, 1980). Empathy has been positively associated to a number of well- being activities, such as volunteering (Batson et al., 1987), charity (Pavey et al., 2012), and longevity (Poulin et al., 2013), and in consumer marketing, advertising and customer interfaces (Wang et al., 2016; Escalas and Stern, 2003). Works on empathy in text have focused on spoken dialogue, addressing conversational agents, psychological interventions, Figure 1: EMPATHBERT architecture. or call center transcripts (McQuiggan and Lester, 2007; Fung et al., 2016; P´erez-Rosas et al., 2017; Alam et al., 2018; Demasi et al., 2019). Buechel et al. (2018) collected an empathy-distress dataset by leveraging users’ reactions to textual stimulus content. Sedoc et al. (2019) constructed an empathy lexicon by obtaining word ratings from document- level ratings from this dataset. Xiao et al. (2012); Gibson et al. (2015); Khanpour et al. (2017) pre- sented predictive models for empathy in the health- care domain. However, we believe none of the above works focus on (a) predicting empathy from textual reactions, and (b) studying the impact of demographics on the expression of empathy. Language preferences vary with user demo- graphics (Tresselt and Mayzner, 1964; Eckert and McConnell-Ginet, 2013; Garimella et al., 2016; Lin et al., 2018; Loveys et al., 2018), and this has led to studies leveraging the user demographic information to obtain better language representa- tions and classification models for various NLP 3073 tasks (Volkova et al., 2013; Bamman et al., 2014; Hovy, 2015; Garimella et al., 2017). Owing …

Content truncated to 500 words