Aspect-based sentiment analysis for German : analyzing 'talk of literature' surrounding literary prizes on social media

Publication type
Publication status
De Greve, L., Singh, P., Van Hee, C., Lefever, E., & Martens, G.
View in Biblio
(externe link)


Since the rise of social media, the authority of traditional professional literary critics has beensupplemented – or undermined, depending on the point of view – by technological developmentsand the emergence of community-driven online layperson literary criticism. So far, relatively littleresearch (Allington 2016, Kellermann et al. 2016, Kellermann and Mehling 2017, Bogaert 2017, Pi-anzola et al. 2020) has examined this layperson user-generated evaluative “talk of literature”instead of addressing traditional forms of consecration. In this paper, we examine the layper-son literary criticism pertaining to a prominent German-language literary award: the Ingeborg-Bachmann-Preis, awarded during the Tage der deutschsprachigen Literatur (TDDL).We propose an aspect-based sentiment analysis (ABSA) approach to discern the evaluativecriteria used to differentiate between ‘good’ and ‘bad’ literature. To this end, we collected a cor-pus of German social media reviews, retrieved from Twitter, and enriched it with manual ABSAannotations:aspectsand aspect categories (e.g. the motifs or themes in a text, the jury discus-sions and evaluations, ...),sentiment expressionsandnamed entities. In a next step, the manualannotations are used as training data for our ABSA pipeline including 1) aspect term categoryprediction and 2) aspect term polarity classification. Each pipeline component is developed usingstate-of-the-art pre-trained BERT models.Two sets of experiments were conducted for the aspect polarity detection: one where only theaspect embeddings were used and another where an additional context window of five adjoiningwords in either direction of the aspect was considered. We present the classification results forthe aspect category and aspect sentiment prediction subtasks for the Twitter corpus. Thesepreliminary experimental results show a good performance for the aspect category classification,with a macro and a weighted F1-score of 69% and 83% for the coarse-grained and 54% and 73% forthe fine-grained task, as well as for the aspect sentiment classification subtask, using an additionalcontext window, with a macro and a weighted F1-score of 70% and 71%, respectively