CantoMap: a Hong Kong Cantonese MapTask Corpus

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2906–2913 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2906 CantoMap: a Hong Kong Cantonese MapTask Corpus Grégoire Winterstein, Carmen Tang, Regine Lai Université du Québec à Montréal, The Chinese University of Hong Kong Département de Linguistique, Department of Linguistics and Modern Languages winterstein.gregoire@uqam.ca, tangkm.carmen@gmail.com, ryklai@cuhk.edu.hk Abstract This work reports on the construction of a corpus of connected spoken Hong Kong Cantonese. The corpus aims at providing an additional resource for the study of modern (Hong Kong) Cantonese and also involves several controlled elicitation tasks which will serve different projects related to the phonology and semantics of Cantonese. The word-segmented corpus offers recordings, phonemic transcription, and Chinese characters transcription. The corpus contains a total of 768 minutes of recordings and transcripts of forty speakers. All the audio material has been aligned at utterance level with the transcriptions, using the ELAN transcription and annotation tool. The controlled elicitation task was based on the design of HCRC MapTask corpus (Anderson et al., 1991), in which participants had to communicate using solely verbal means as eye contact was restricted. In this paper, we outline the design of the maps and their landmarks and the basic segmentation principles of the data and various transcription conventions we adopted. We also compare the contents of CantoMap to other available Cantonese corpora. Keywords: Hong Kong Cantonese, corpus, interaction, MapTask, tone merging, word segmentation 1. Introduction This paper introduces CantoMap, an audio and transcribed corpus of connected speech in Hong Kong Cantonese. Comparable previous works on Cantonese corpora include The Hong Kong Cantonese Adult Language Corpus (HK- CAC, (Leung and Law, 2001) ), The Hong Kong Can- tonese Corpus (HKCanCor, (Luke and Wong, 2015) ), the Linguistics Corpus of Mid-20th Century Hong Kong Can- tonese (Chin, 2015) and the recordings of Adult-Child ses- sions from CHILDES (Lee et al., 1994; Fletcher et al., 2000). Some of these corpora were phonetically transcribed with IPA, while others were phonemically transcribed and glossed with parts of speech, and a subset of these were transcribed with Chinese characters. Some were manually segmented, others were semi-automatically segmented and some are not segmented at all. Thus, although each of these resources is precious in its own right, none of them is tran- scribed in the same way, and each lacks information that appears in the others. In addition, at the present moment, none of these corpora offers access to the recordings of the conversation transcribed in the form of quality audio files. The size of each corpus is also rather limited, and to date, there is still a lack of resources to apply data-intensive tech- niques of NLP to Hong Kong Cantonese. Addressing those issues is one of the goals of CantoMap, i.e. adding to the list of existing resources and offer a corpus that brings all levels of analysis in a single resource: access to the audio material as well as to aligned transcriptions at …

Content truncated to 500 words