The ManDi Corpus: A Spoken Corpus of Mandarin Regional Dialects

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1985–1990 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 1985 The ManDi Corpus: A Spoken Corpus of Mandarin Regional Dialects Liang Zhao1 and Eleanor Chodroff1 University of York1 YO10 5DD, Heslington, York, UK liang.zhao@york.ac.uk, eleanor.chodroff@york.ac.uk Abstract In the present paper, we introduce the ManDi Corpus, a spoken corpus of regional Mandarin dialects and Standard Mandarin. The current version of the corpus contains 357 recordings (about 9.6 hours) of monosyllabic words, disyllabic words, short sentences, a short passage and a poem, each produced in Standard Mandarin and in one of six regional Mandarin dialects: Beijing, Chengdu, Jinan, Taiyuan, Wuhan, and Xi’an Mandarin from 36 speakers. The corpus was collected remotely using participant-controlled smartphone recording apps. Word- and phone-level alignments were generated using Praat and the Montreal Forced Aligner. The pilot study of dialect-specific tone systems showed that with practicable design and decent recording quality, remotely collected speech data can be suitable for analysis of relative patterns in acoustic-phonetic realization. The corpus is available on OSF (https://osf.io/fgv4w/) for non-commercial use under a CC BY-NC 3.0 license. Keywords: speech, Mandarin Chinese dialects, remote data collection, smartphone recordings 1. Introduction Speech science has traditionally relied on high quality speech recordings, collected in person with a standardized recording setup in laboratories/studios or in field settings. The strictly controlled environment for speech recording allows the acoustic-phonetic analysis to be focused on differences in the speech alone. However, recent global developments ranging from COVID-19 to climate change have triggered a comprehensive re-evaluation of our approach to research and travel, while researchers have drastically rethought data collection approaches and turned to varieties of remote speech data collection. Remote audio collection hereby refers to the scenario where the researcher delivers an experiment to a participant virtually and remotely, who in turn records his/her speech using a personally available recording device, such as a smartphone or computer, and sends the recording back to the experimenter. Early attempts at remote audio collection were inevitably confined to the state of technological development at the time. Previous methods in the 1990s exploited access to landline telephones to collect speech corpora (e.g. CALLHOME, Canavan and Zipperlen, 1996). This method increased the range of participant enrolment, but the recordings were limited by the telephone bandwidth, and were thus of low quality for phonetic research. Thanks to current smartphone technology, high quality recording baseline and its widespread accessibility can advance remote audio collection for research in basic and forensic speech sciences, and speech engineering. As approximately 50% of the world’s population is estimated to own a smartphone and this figure is likely to rise, remote audio collection with smartphone devices could increase both the amount and range of speech data. With a robust method, we can simultaneously reach a vast number of diverse communities around the world: anyone with access to a smartphone could contribute. Speech technologies such as speech-to-text systems and text-to-speech synthesis might also benefit from …

Content truncated to 500 words