New PERL model enhances Chinese ASR error correction with phonetic and semantic fusion

Researchers have developed PERL, a new language model designed to correct errors in Chinese Automatic Speech Recognition (ASR) systems. PERL addresses the challenge of phonetic errors and length constraints by employing a constrained rephrasing pipeline. This model fuses semantic and phonetic (Pinyin) representations, and enforces target length through mask budgeting, leading to significant reductions in character error rates on benchmarks like Aishell-1 and DoAD. AI

IMPACT Improves accuracy for Chinese speech recognition systems, potentially aiding downstream applications.

RANK_REASON The cluster contains a research paper detailing a new model for a specific NLP task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• Aishell-1
• Junhong Liang
• PERL
• pinyin

• paper
• model release

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