Joyo Kanji Yomi Benchmark

A kanji-level pronunciation evaluation benchmark for Japanese TTS, covering all 2,136 Joyo kanji and their 4,378 readings with 13,095 native-speaker-verified test sentences.

Dataset Description

Each sample targets a specific kanji-reading pair. The sentence context is designed so that only the target reading is valid. All sentences and annotations have been verified by 35 native Japanese speakers through a three-stage review process.

A small number of kanji-reading pairs are excluded when the target reading cannot be uniquely disambiguated from the kanji's other readings by sentence context alone.

Data Format

Each sample is a JSON object with the following fields:

Example

{
  "key": "精_セイ_0",
  "normalized_text": "その時計は非常に精密な作りになっている。",
  "normalized_pron": "ソノトケーワヒジョーニ<セー>ミツナツクリニナッテール。"
}

In normalized_pron, the <セー> marks the katakana substring corresponding to the target kanji 「精」, enabling automatic extraction for kanji-level CER computation.

Usage

For evaluation, use the evaluation toolkit which handles ASR transcription, alignment, and metric computation.

Citation

@misc{liu2026sarashina22ttstacklingkanjipolyphony,
      title={Sarashina2.2-TTS: Tackling Kanji Polyphony in Japanese Speech Generation via Data Scaling and Targeted Data Synthesis}, 
      author={Lianbo Liu and Shiao Zhu and Kai Washizaki and Reo Yoneyama and Haesung Jeon and Mengjie Zhao and Yusuke Fujita and Hao Shi and Nao Yoshida and Yuan Gao and Roman Koshkin and Yukiya Hono and Yui Sudo},
      year={2026},
      eprint={2606.25369},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2606.25369}, 
}

License

This project is licensed under the MIT License.