A Corpus of Turkish Offensive Language on Social Media

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 600–609 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 600 RDG-Map: A Multimodal Corpus of Pedagogical Human-Agent Spoken Interactions Maike Paetzel1, Deepthi Karkada2, Ramesh Manuvinakurike3 1Uppsala University Social Robotics Lab, Sweden 2Intel Corp, 3Intel Labs, Hillsboro, Oregon, United States maike.paetzel@it.uu.se, deepthi.karkada@intel.com, ramesh.manuvinakurike@intel.com Abstract This paper presents a multimodal corpus of 209 spoken game dialogues between a human and a remote-controlled artificial agent. The interactions involve people collaborating with the agent to identify countries on the world map as quickly as possible, which allows studying rapid and spontaneous dialogue with complex anaphoras, disfluent utterances and incorrect descriptions. The corpus consists of two parts: 8 hours of game interactions have been collected with a virtual unembodied agent online and 26.8 hours have been recorded with a physically embodied robot in a research lab. In addition to spoken audio recordings available for both parts, camera recordings and skeleton-, facial expression- and eye-gaze tracking data have been collected for the lab-based part of the corpus. In this paper, we introduce the pedagogical reference resolution game (RDG-Map) and the characteristics of the corpus collected. We also present an annotation scheme we developed in order to study the dialogue strategies utilized by the players. Based on a subset of 330 minutes of interactions annotated so far, we discuss initial insights into these strategies as well as the potential of the corpus for future research. Keywords: Multimodal Corpus, Conversational Games, Serious Dialogue Games, Rapid Spoken Dialogue, Human-Robot Inter- action, Crowd-sourcing 1. Introduction In this paper, we present the RDG-Map corpus, an exten- sion of the previously developed Rapid Dialogue Game (RDG) series (Paetzel et al., 2014; Zarrieß et al., 2016). The core dynamic of the RDG domain is to score as high as possible in a spoken dialogue game under time pressure, which leads to spontaneous rapid responses from the play- ers. Due to its importance in daily conversations, the goal of rapid dialogue games is to capture related phenomena of spontaneous natural speech in numerous tasks of vary- ing complexity. Such corpora can then be used for building state-of-the-art dialogue processing models that can make dialogue systems more efficient and human-like (Paetzel et al., 2015; Manuvinakurike et al., 2017). In comparison to the other games developed as part of the RDG series, RDG-Image, RDG-Phrase, and RDG-Pento, finding countries on the world map requires a more com- plex description strategy than it is required for identifying images, phrases and Pentomino pieces. In these simpler do- mains, players followed straightforward dialogue strategies only involving the current target. In RDG-Map, on the con- trary, we observe more complex dialogue interactions mak- ing use of other countries and past context. Such strategies result in rich conversations that can be an ideal platform for research on a variety of dialogue-related topics, such as language understanding, generation, and dialogue manage- ment. Figure 1 shows a sample dialogue in this domain. The RDG-Map game contributes …

Content truncated to 500 words