New models improve phonetic alignment for Chengdu Mandarin

Researchers have developed new phonetic forced alignment models specifically for low-resource language varieties, focusing on Chengdu Mandarin. They trained a text-dependent GMM-HMM model, Chengdu-MFA, and a text-independent model, Chengdu-FC, using a 17-hour corpus. Both models demonstrated significant improvements over Standard Mandarin baselines, with Chengdu-MFA reducing phone boundary differences by 31.8% and Chengdu-FC achieving a 61.2% reduction, establishing a pipeline for developing aligners without extensive manual annotation. AI

IMPACT Establishes a practical pipeline for developing accurate phonetic aligners for under-resourced language varieties.

RANK_REASON The cluster contains an academic paper detailing new model training and evaluation for a specific language variety. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Chengdu Mandarin
• Chengdu-MFA
• GMM-HMM
• Hugging Face

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →