eduhk-compling/11557553_ChanTakChing

Dataset Description

This dataset provides a reading of the Wikipedia page of Metamorphosis in Cantonese, with slight alterations. The text is spoken in written language, and not the cadence of spoken language. The audio files were recorded through the recorder app on a Xiaomi phone in a room at home from 7:28pm to 7:58pm. I, the speaker, am a young adult male born-and-raised in Hong Kong.

Issues Encountered & Solution

I encountered the issue of some words having multiple pronunciations, such as 身, which had 4 different pronunciations attributed to it. Hence, in order to assure accuracy in transcription, I opted to use the Multi-function Chinese Character Database, and manually transcribed every syllable. The process was tedious, hence I would reuse transcriptions I had already researched before in excerpts with identical pronunciation, negating the need to find every syllable.

Another issue I encountered was that the recording picked up particularily loud exhalations, which I had resolved by cutting the section out, as these gaps can only be created when I am not speaking. There was no other background noise as I had ensured the room was as silent as possible before recording.

Sometimes the dataset viewer is not available for the split on the dataset card.

Server error while post-processing the split rows. Please report the issue.

Error code: RowsPostProcessingError

To view the dataset, try the Data Studio.