Datasets:
language:
- zh
- en
tags:
- tts
- bilingual
- common-sentences
license: cc-by-4.0
task_categories:
- text-to-speech
Bilingual Common Sentence Dataset for TTS
Dataset Description
This dataset is a bilingual dataset consisting of commonly used sentences in Chinese and English, designed for Text-to-Speech (TTS) research and learning purposes. It contains short, sentence-level audio recordings of everyday sentences spoken by a single speaker in a quiet environment. Each audio file is provided in WAV format and is paired with its corresponding transcription in the metadata file. For Chinese sentences, phonetic transcriptions in pinyin are also included. The dataset is intended for educational use, especially for understanding how common sentences are prepared and structured for TTS systems.
Reflection: Issues Encountered and Solutions
One challenge encountered during dataset preparation was maintaining consistent pronunciation when switching between Chinese and English common sentences. To address this issue, all recordings were completed in a single session, and unclear sentences were re-recorded when necessary. Another difficulty was segmenting long recordings into sentence-level audio clips while ensuring accurate alignment between audio and text. This was resolved by carefully reviewing each WAV file and cross-checking it with the metadata. These steps helped ensure the dataset is clean and suitable for TTS applications.