UgChDial: A Uyghur Chat-based Dialogue Corpus for Response Space Classification

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 3140–3149 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 3140 UgChDial: A Uyghur Chat-based Dialogue Corpus for Response Space Classification Zulipiye Yusupujiang, Jonathan Ginzburg Universit´e Paris Cit´e, CNRS, Laboratoire de Linguistique Formelle zulipiye.yusupujiang@linguist.univ-paris-diderot.fr, yonatan.ginzburg@u-paris.fr Abstract In this paper, we introduce a carefully designed and collected language resource: UgChDial – a Uyghur dialogue corpus based on a chatroom environment. The Uyghur Chat-based Dialogue Corpus (UgChDial) is divided into two parts: (1). Two-party dialogues and (2). Multi-party dialogues. We ran a series of 25, 120-minutes each, two-party chat sessions, totaling 7323 turns and 1581 question-response pairs. We created 16 different scenarios and topics to gather these two-party conversations. The multi-party conversations were compiled from chitchats in general channels as well as free chats in topic-oriented public channels, yielding 5588 unique turns and 838 question-response pairs. The initial purpose of this corpus is to study query-response pairs in Uyghur, building on an existing fine-grained response space taxonomy for English. We provide here initial annotation results on the Uyghur response space classification task using UgChDial. Keywords: Dialogue Corpus, Response Space Classification, Uyghur, Less-Resourced Languages 1. Introduction Characterizing the response space of questions is of great importance, as it is a critical element in devel- oping a dialogue system that interacts in a natural way. It also provides a fundamental benchmark for dialogue/question theories. Building on detailed cor- pus analysis in English (the British National Corpus (Burnard, 2007) and three other more genre-specific corpora (BEE (Ros´e et al., 1999), MapTask (Anderson et al., 1991), Cornell Movie (Danescu-Niculescu-Mizil and Lee, 2011))) and in Polish (Pezik, 2014), as well as on formal dialogue semantic analysis in the framework of KoS (Ginzburg, 2012; Łupkowski and Ginzburg, 2016), (Ginzburg et al., 2019; Ginzburg et al., 2022) propose that the class of responses to a question q1 can be partitioned into three main categories: (1) a. q(uestion)–specific: responses directly or in- directly about or subquestions of q1; b. MetaCommunicative: responses directly about or subquestions of a question defined in part from the utterance of q1; c. Evasion: responses directly about or subques- tions of a question that is distinct from q1 and arises from some other component of the con- text: 1. Ignore (address the situation, but not the question; e.g., Anon: on the Sunday be- fore you killed the animals, you didn’t in fact feed them. Why was that? Harry: Only water. (BNC)); 2. Change the topic (e.g., Nicola: Come on, let’s get dressed. Which pants are you wearing? Oliver: What’s he got on his mouth? (BNC)); 3. Motive (‘Why do you ask?’); 4. Difficult to provide a response (‘I don’t know’). The taxonomy is introduced in detail in Section 5. In light of these studies, we aim to address the chal- lenge of characterizing the response space to questions in a low-resource language – Uyghur. The Uyghurs are Turkic ethnic groups native to the Xinjiang Uyghur Autonomous Region in Northwest …

Content truncated to 500 words