| license: apache-2.0 | |
| language: | |
| - ja | |
| tags: | |
| - mos | |
| - speech-quality-assessment | |
| - pitch-accent | |
| - wav2vec2 | |
| - audio | |
| # PASQA — Pitch-Accent-Focused Speech Quality Assessment Model | |
| Pretrained checkpoint for **[PASQA](https://github.com/lycorp-jp/PASQA)**, an SSL-based MOS prediction model | |
| that estimates the perceptual quality of synthesized Japanese speech with a focus on | |
| **pitch-accent correctness**. | |
| The model is developed by LY Corporation and released alongside our INTERSPEECH 2026 paper. | |
| ## Usage | |
| Inference is done with the code in the [PASQA](https://github.com/lycorp-jp/PASQA) repository. See the repository for setup and usage instructions. | |
| ## License | |
| These model parameters are released under the | |
| [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0). | |