Infore2 Audiobooks: 415 Hours of Vietnamese Narration for Chinese Wuxia and Xianxia

Infore2 Audiobooks is an unofficial mirror of InfoRe Technology's public dataset number 2, containing Vietnamese audiobook narrations of Chinese wuxia and xianxia fiction. The dataset comprises 315,000 audio samples totaling 415 hours, extracted from YouTube and automatically labeled using text alignment techniques. It was created by doof-ferb and last updated on Hugging Face in February 2024.

Use Cases

• Train Vietnamese automatic speech recognition models based on the 415 hours of labeled audio.
• Develop text-to-speech synthesis for Vietnamese using the aligned audio and text data.
• Research audio-text alignment techniques using the automatically labeled samples.
• Build models for genre-specific (wuxia/xianxia) speech processing based on the described content.

Strengths

• Contains 315,000 audio samples, providing a substantial volume of data.
• Totals 415 hours of Vietnamese speech, a significant duration for model training.
• Focuses on a specific genre (Chinese wuxia/xianxia), offering domain-specific content.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

InfoRe Technology (unofficial mirror by doof-ferb)

Collection Method

Extracted from YouTube readings of wuxia and xianxia stories, with automatic labeling via text alignment.

Freshness

Last updated 2024-02-11 18:43:29; freshness should be verified.