x-enVENT: A Corpus of Event Descriptions with Experiencer-specific Emotion and Appraisal Annotations

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1365–1375 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 1365 x-enVENT: A Corpus of Event Descriptions with Experiencer-specific Emotion and Appraisal Annotations Enrica Troiano∗, Laura Oberl¨ander∗, Maximilian Wegge, Roman Klinger Institut f¨ur Maschinelle Sprachverarbeitung University of Stuttgart {enrica.troiano,laura.oberlaender,maximilian.wegge,roman.klinger}@ims.uni-stuttgart.de Abstract Emotion classification is often formulated as the task to categorize texts into a predefined set of emotion classes. So far, this task has been the recognition of the emotion of writers and readers, as well as that of entities mentioned in the text. We argue that a classification setup for emotion analysis should be performed in an integrated manner, including the different semantic roles that participate in an emotion episode. Based on appraisal theories in psychology, which treat emotions as reactions to events, we compile an English corpus of written event descriptions. The descriptions depict emotion-eliciting circumstances, and they contain mentions of people who responded emotionally. We annotate all experiencers, including the original author, with the emotions they likely felt. In addition, we link them to the event they found salient (which can be different for different experiencers in a text) by annotating event properties, or appraisals (e.g., the perceived event undesirability, the uncertainty of its outcome). Our analysis reveals patterns in the co-occurrence of people’s emotions in interaction. Hence, this richly-annotated resource provides useful data to study emotions and event evaluations from the perspective of different roles, and it enables the development of experiencer-specific emotion and appraisal classification systems. Keywords: emotion analysis, corpus, affective computing, role labeling, emotion experiencer, appraisal theories, events 1. Introduction Computational emotion analysis from text includes var- ious subtasks, with the (arguably) most popular one being emotion classification – i.e., to categorize texts into a predefined set of emotion classes (Mohammad et al., 2018). The adopted categories often coincide with the list of “basic emotions” proposed by Ekman (1992), namely fear, joy, sadness, anger, disgust, surprise, or with inventories defined by other theorists, like Plutchik (2001). In addition to discrete labels, some works have been identifying structured information, namely emo- tion roles, which divides texts into spans that correspond to semantic roles. Depending on the considered domain, it can aim at distinguishing an emotion experiencer, a stimulus (i.e., the triggering event), a target towards which the emotion is directed, or a cue word evoking a specific affective state (Mohammad et al., 2014). Structured knowledge of this sort has been proven in- formative for emotion classification (Oberl¨ander et al., 2020), but to face one challenge at a time, the two strands of research are typically addressed separately. Hence, when the emotion experiencer is not known to a classification model, a preliminary decision has to be made: the classification should regard either the emo- tion expressed by the writer (Mohammad, 2012, i.a.), or the one triggered in the reader of the text (Haider et al., 2020, i.a.). Studies that include both perspectives are rare (Bostan et al., 2020; Buechel and …

Content truncated to 500 words