ELAL: An Emotion Lexicon for the Analysis of Alsatian Theatre Plays

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5001–5010 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5001 ELAL: An Emotion Lexicon for the Analysis of Alsatian Theatre Plays Delphine Bernhard, Pablo Ruiz Fabo Université de Strasbourg, LiLPa UR 1339 F-67000 Strasbourg, France {dbernhard, ruizfabo}@unistra.fr Abstract In this work, we present a novel and manually corrected emotion lexicon for the Alsatian dialects, including graphical variants of Alsatian lexical items. These High German dialects are spoken in the North-East of France. They are used mainly orally, and thus lack a stable and consensual spelling convention. There has nevertheless been a continuous literary production since the middle of the 17th century and, in particular, theatre plays. A large sample of Alsatian theatre plays is currently being encoded according to the Text Encoding Initiative (TEI) Guidelines. The emotion lexicon will be used to perform automatic emotion analysis in this corpus of theatre plays. We used a graph-based approach to deriving emotion scores and translations, relying only on bilingual lexicons, cognates and spelling variants. The source lexicons for emotion scores are the NRC Valence Arousal and Dominance and NRC Emotion Intensity lexicons. Keywords: emotion, lexicon, Alsatian, theatre 1. Introduction Automatic emotion analysis for written corpora of the- atre plays has been little investigated to date (see Sec- tion 2.3). This is nevertheless a promising avenue, since previous work based on novels has shown that the characters’ direct speech contains salient emotional features (Henny-Krahmer, 2018). There are two ways in which emotions can be expressed in theatre plays: (i) Directly in characters’ speech, through emotion- bearing words; (ii) Indirectly in stage directions, which indicate how a character should behave or look like on stage. For instance, Kim and Klinger (2019) have anal- ysed a corpus of fan fiction to show how emotions are also expressed non-verbally, e.g., facial expressions, gaze, voice characteristics, gestures, body postures. Sentiment lexicons are an important resource for au- tomatic emotion analysis. While expressions in sen- timent lexicons should be interpreted with respect to contextual cues (negations, long distance dependen- cies, ambiguity), lexicon-based approaches are never- theless easier to implement in a low-resource setting, where annotated data amenable to supervised learning are notably lacking. In this work, we focus on the Alsatian dialects, High German dialects spoken in the North-East of France. These dialects are used mainly orally, and thus lack a stable and consensual spelling convention. There has nevertheless been a continuous literary production since the middle of the 17th century and, in particular, theatre plays. Recently, image-mode digital versions for a large sample of theatre plays, rep- resentative of the tradition, were created by the BNU.1 Within the MeThAl project (Ruiz Fabo et al., 2021),2 1The BNU is Strasbourg’s University and National Li- brary. The plays are at https://numistral.fr/fr/ theatre-alsacien, accessible via the Découvrir (i.e. Discover) link. 2https://methal.pages.unistra.fr we are currently performing Optical Character Recog- nition (OCR) on this corpus and encoding it accord- ing to the Text Encoding …

Content truncated to 500 words