Dataset Card for Nexdata/Thai_Children_Spontaneous_Speech_Data

Description

The 100 Hours - Thai Child's Spontaneous Speech Data, manually screened and processed. Annotation contains transcription text, speaker identification, gender and other informantion. This dataset can be applied in speech recognition (acoustic model or language model training), caption generation, voice content moderation and other AI algorithm research.

For more details, please refer to the link: https://www.nexdata.ai/datasets/speechrecog/1330?source=Huggingface

Specifications

Format

16k Hz, 16 bit, wav, mono channel;

Age

12 years old and younger children;

Content category

including self-media, conversation, live, lecture, variety show;

Language

Thai;

Annotation

annotation for the transcription text, speaker identification, gender;

Accuracy

Word Accuracy Rate (WAR) at least 98%.

Licensing Information

Commercial License