A Multi-Party Dialogue Ressource in French

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 814–823 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 814 A Multi-Party Dialogue Ressource in French Maria Boritchev, Maxime Amblard Institute of Mathematics of the Polish Academy of Sciences Warsaw, Poland LORIA, UMR 7503, Université de Lorraine, CNRS, Inria, 54000 Nancy, France mboritchev@impan.pl, maxime.amblard@loria.fr Abstract We present Dialogues in Games (DinG), a corpus of manual transcriptions of real-life, oral, spontaneous multi-party dialogues between French-speaking players of the board game Catan. Our objective is to make available a quality resource for French, composed of long dialogues, to facilitate their study in the style of (Asher et al., 2016). In a general dialogue setting, participants share personal information, which makes it impossible to disseminate the resource freely and openly. In DinG, the attention of the participants is focused on the game, which prevents them from talking about themselves. In addition, we are conducting a study on the nature of the questions in dialogue, through annotation (Cruz Blandon et al., 2019), in order to develop more natural automatic dialogue systems. Keywords: Dialogue, Transcription, Annotation, Questions 1. Introduction We envision our corpus as a corpus of spontaneous dia- logues in French with a quality transcription. Its nature allows for large dissemination and high cross-domain reusability. Its length allows for a study from differ- ent perspectives. As the cost of producing resources is very high, we designed our corpus aiming for the widest possible use. Furthermore, we want to present it as a corpus that follows the good practices of corpus collection, usable for the collection of other corpora, especially ones that cannot be disseminated largely be- cause of the nature of the data they contain. One of the main inspirations in the design of DinG was the STAC project1 corpus (Asher et al., 2016). The cor- pus of STAC is composed of chat logs from an online version of the board game Catan, played by English speakers. STAC is annotated in Segmented Discourse Representation Theory (SDRT, (Asher and Lascarides, 2003)). The proximity our two corpora share is a great opportunity to progress on comparisons between writ- ten and oral discourse on the same topics while opening on dialogue. DinG is now used to study questions in dialogue in approaches such as (Boritchev and Amblard, 2020; Boritchev, 2021). As DinG is composed of long human-human interactions, it can be very largely used for dialogue studies in numerous fields. Links with other Studies of Dialogue Another mo- tivation for developing DinG is to showcase the work we are conducting on highly sensible health-related data. Studying DinG can help not to tag a phenomenon as a conversational disorder by acknowledging the fact that this phenomenon occurs in a non-pathological set- ting as well. Data from DinG is also used in the Séma- gramme team (LORIA, Inria Nancy Grand-Est) to help 1https://www.irit.fr/STAC/ develop and showcase SLODiM2, a tool for dialogue analysis in a medical (psychiatric) setting, developed in the line …

Content truncated to 500 words