Zhang Xuefeng Conversational Speech Corpus for AI Style Transfer

A corpus of conversational speech transcripts from Zhang Xuefeng's live-streamed content on college admissions, processed for AI training. The dataset was created by user 'england-lobster' and last updated on June 13, 2026. It is derived from automated speech recognition of public broadcasts, followed by multi-stage cleaning and episode construction.

Use Cases

• Fine-tuning large language models for style mimicry based on described conversational patterns.
• Training digital human avatars for live-streaming based on the dialogue structure.
• Developing Retrieval-Augmented Generation (RAG) systems with a consistent persona voice.
• Studying informal Chinese speech patterns and rhetorical devices in educational live streams.

Strengths

• Explicitly designed for style fidelity, preserving oral speech patterns like filler words and rhetorical repetitions.
• Underwent multi-stage cleaning for pathological stutters and ASR homophone errors.
• Built for long-range conversational coherence, a key feature for live-streaming avatars.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count, column definitions, and file formats are unknown, which may limit suitability assessment.
• Data is explicitly not factually accurate, intended only for style mimicry.

Provenance

Source

Public live-stream and call-in content from Zhang Xuefeng.

Collection Method

Automated Speech Recognition (ASR) followed by LLM-assisted multi-stage cleaning and episode construction.

Freshness

Last updated 2026-06-13 14:57:03; freshness should be verified.