Emotion Experiencer Recognition as a Prerequisite for Experiencer-Specific Emotion Analysis

Emotion role labeling aims at extracting who is described in text to experience an emotion, why, and towards whom. This is often a challenging modelling task which might be overly sophisticated if the main question to answer is who feels which emotion. Recently, Troiano et al. (2022) proposed a data set that focuses on assigning emotion labels and appraisal labels to individual entities in text and Wegge et al. (2022) presented the first modelling experiments. Their experiencer-specific emotion prediction model has, however, only been evaluated on gold-annotated experiencers, due to the unavailability of an automatic experiencer detection approach. We fill this gap with the first experiments to automatically detect emotion experiencers in text and, subsequently, assign them emotions. We show that experiencer detection in text is a challenging task, with a precision of .82 and a recall of .56 (F1 =.66). Consequently, the performance of the experiencer-specific emotion detection pipeline drops with these predictions in comparison to using gold experiencer annotations. This motivates future work of jointly modelling emotion experiencer detection and emotion/appraisal recognition.