Tai Lo To Chinese Parallel Text Corpus

Over 10,000 example sentences from the MoeDict-cmn-hak-10k dataset contribute to this collection of Chinese text paired with its Tai-lo transliteration. Curiousfox compiled this dataset, last updated in January 2026, using transliteration tools from taibun.andreihar.com and ÌTHUÂNKHOKI.

Use Cases

• Train a sequence-to-sequence model to transliterate Chinese characters to Tai-lo romanization using the parallel sentence pairs.
• Fine-tune a bilingual language model for Hokkien using the Chinese-Tai-lo text pairs as a parallel corpus.
• Analyze the phonetic mapping rules between Chinese characters and Tai-lo romanization by comparing the source and target text fields.
• Develop a transliteration evaluation benchmark for Hokkien using the provided human-curated example sentences.

Strengths

• Integrates over 10,000 sentences from the MoeDict-cmn-hak-10k dataset
• Leverages specialized transliteration tools from taibun.andreihar.com and ÌTHUÂNKHOKI

Limitations

• Unknown total row count and dataset size
• Unknown geographic distribution of the Hokkien dialect variants covered

Provenance

Source

Aggregated from Jonshon8187/Chinese_Multi-Emotion_Dialogue_Dataset, RicardoRei/wmt-mqm-human-evaluation, Ayaka/MoeDict-cmn-hak-10k, sarahwei/Taiwanese-Minnan-Sutiau, sarahwei/Taiwanese-Minnan-Example-Sentences

Collection Method

Transliteration of Chinese source text to Tai-lo using external tools

Freshness

Last updated January 2026