Tran-Vi-Teacher: Chinese-Vietnamese Web-Novel Parallel Corpus

350,751 rows of Chinese-to-Vietnamese parallel text for web-novel translation training. The dataset was created by author ngocdang83 using Gemini 2.5/3.0/3.1 teacher models to distill raw Chinese web-novel paragraphs. It was last updated on 2026-05-22.

Use Cases

• Train machine translation models based on the described Chinese-Vietnamese parallel corpus.
• Fine-tune translation systems for web-novel genres like xianxia, urban, and fantasy.
• Evaluate translation quality for multi-line paragraph chunks, which the description mentions as a format.
• Develop domain-specific translation tools for literary content.
• Conduct linguistic analysis on translated web-novel text across different sub-genres.

Strengths

• Contains 350,751 rows of parallel text, providing a substantial corpus for training.
• Described as a 'strict-clean' parallel corpus, suggesting high-quality alignment.
• Covers multiple web-novel domains including xianxia, urban, school, fantasy, history, and sci-fi.
• Uses paragraphs with a median of 7 lines per row, providing context for translation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known but total file size and specific file formats are unknown.
• The dataset may reflect stylistic and thematic biases inherent to its source of Chinese web-novels.

Provenance

Source

Distilled from Gemini 2.5/3.0/3.1 teacher models on raw Chinese web-novel paragraphs.

Collection Method

Likely involves automated distillation and cleaning processes from teacher model outputs.

Freshness

Last updated 2026-05-22 16:52:57; freshness should be verified.