Researchers have developed MERaLiON-GR, a novel speech gender recognition model capable of classifying gender for both English and several Southeast Asian languages. This model is built upon MERaLiON-SpeechEncoder-2, a conformer-based transformer, and utilizes Low-Rank Adaptation (LoRA) for efficient fine-tuning. MERaLiON-GR demonstrates superior performance compared to existing state-of-the-art models like Vox-Profile and a large Audio-LLM across various evaluation modes and languages, including Chinese, Malay, Tamil, Thai, Vietnamese, Indonesian, and Khmer. AI
IMPACT This research advances multilingual speech processing capabilities, potentially improving accessibility and understanding across diverse linguistic communities.
RANK_REASON The item is an academic paper detailing a new model and its performance. [lever_c_demoted from research: ic=1 ai=1.0]
- Audio-LLM
- ECAPA-TDNN
- English
- Indonesian
- Khmer
- Lora
- Malay
- MERaLiON-GR
- MERaLiON-SpeechEncoder-2
- Standard Chinese
- Tamil
- Thailand
- Vietnamese
- Vox-Profile
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →