Datasets:
license: cc-by-4.0
language:
- en
tags:
- audio
- speech
Dataset Description
This dataset is designed for reading out stories. Each sentence describes a clear scene with a setting, characters, and an action, so when read aloud in sequence, they resemble a series of very short story moments rather than isolated phrases. This makes the data suitable for studying how readers produce connected narrative-style speech, including prosody, intonation, and timing when telling simple stories about everyday situations.
Issues Encountered & Solution
I encountered two problems. The first probelm was that I find transforming plain text file to csv file technically challenging for me. Unfortunately I seldom use the csv format in my daliy life, so I spend some time to explore ways in order to fulfill the requirement. At the end my solution was uploading a csv file from one of my friend, and then changing the content of the file directly on hugging face. I struggled to balance natural rhythm with careful pronunciation. Over‑articulating sounded robotic, but relaxed reading produced more reductions and linking. This tension made me reflect on speaking style choices and pushed me to aim for a “careful but natural” pronunciation that is consistent and clearly documented.