ToneCL improves few-shot syllable tone recognition in low-resource languages
ToneCL: Contrastive Learning for Few-Shot Syllable-Level Tone Classification
Computation and Language
Summary
Many languages use tones to change the meaning of words, but most have little data available to teach computers to understand these tones. The authors designed ToneCL, a method that learns from just a few examples to recognize tones at the level of individual syllables. It first trains on unlabeled speech while focusing on features important for tones, then quickly adapts to new data with very few examples. Their tests on Mandarin and Vietnamese showed that ToneCL works better than existing methods, even when only a handful of labeled examples are available.
What this means in practice
- •For field linguists: Generate accurate syllable-level tone annotations from limited recordings for low-resource languages.
- •For speech technology developers: Build speech recognition systems that handle tone identification effectively with only a few labeled samples per tone class.$Commercial implications: Enables development of tone-aware speech products for low-resource languages, improving user experience in tone-dependent language applications.
Authors
Qisheng Liao, Youngah Do
Abstract
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.