Search papers, labs, and topics across Lattice.
This paper presents GigaAM Multilingual, a foundation model specifically designed to improve automatic speech recognition (ASR) for underrepresented Central Asian languages by leveraging a Conformer encoder pre-trained on 2 million hours of audio. The authors introduce innovative strategies such as cluster-level data balancing during pre-training and domain-aware sampling during fine-tuning to counteract the dominance of head languages. Experimental results show that GigaAM Multilingual significantly outperforms existing models like Whisper Large v3 and Omnilingual-1B on spontaneous speech tasks, demonstrating its effectiveness in addressing data scarcity issues in multilingual ASR.
GigaAM Multilingual achieves remarkable ASR performance improvements for underrepresented languages, outperforming leading models despite severe data limitations.
Despite recent scaling successes, multilingual ASR performance remains highly uneven, with long-tail languages suffering from severe data scarcity. This work addresses the challenge of building robust foundation models for underrepresented Central Asian languages (Kazakh, Kyrgyz, Uzbek). We present GigaAM Multilingual, a Conformer encoder pre-trained on 2M hours of audio using a HuBERT-style objective. Crucially, we introduce a cluster-level data balancing strategy during pre-training and a domain-aware sampling method during fine-tuning to mitigate head-language dominance. In controlled comparisons, our approach outperforms strong open pretrained encoders (Whisper Large v3, Omnilingual-1B) on target languages, achieving significant gains on spontaneous speech while maintaining efficiency. We release the foundation encoder and ASR model, offering a proven recipe for effective multilingual adaptation under realistic data imbalance.
