ToneCL: Contrastive Learning for Few-Shot Syllable-Level Tone Classification
ToneCL reaches 91.6% few-shot Mandarin tone accuracy using contrastive pretraining on unlabeled speech.
ToneCL is a lightweight contrastive-learning method for few-shot syllable-level tone classification in low-resource tone languages. It is pretrained on unlabeled speech with augmentations that preserve tonal identity, then fine-tuned on tens of examples per class using Mandarin and Vietnamese. On six-speaker Mandarin it reaches 91.6% at 10 shots, and Vietnamese pretraining followed by Mandarin fine-tuning reaches 91.0%. Ablations identify frequency-band rejection as the most important augmentation.
- ToneCL targets few-shot syllable-level tone classification for low-resource languages.
- Pretraining uses unlabeled speech with tone-preserving augmentations.
- Six-speaker Mandarin reaches 91.6% accuracy at 10 shots.
- Vietnamese-to-Mandarin transfer reaches 91.0% at 10 shots.
Full article133 words · extracted from arxiv.org · click to collapse
Tone languages constitute over 50-70% of the world's languages, but the vast majority are low-resource, lacking the large transcribed corpora needed for automatic tone classification. Existing datasets are typically collected at the sentence level, whereas field linguists require fine-grained syllable-level annotations. We propose ToneCL, a lightweight contrastive learning framework for few-shot syllable-level tone classification. We simulate low-resource conditions on Mandarin and Vietnamese, limiting labeled data to tens of examples per tone class. ToneCL is pretrained on unlabeled speech with augmentations that preserve tonal identity, then fine-tuned on few-shot examples. Experiments show our method consistently outperforms baselines, achieving 91.6% on six-speaker Mandarin at 10 shots. Cross-lingual transfer is also effective: pretraining on Vietnamese and fine-tuning on Mandarin reaches 91.0\% accuracy at 10 shots. Ablation confirms that frequency band rejection is the most critical augmentation.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.24903