New PAREDA dataset targets ASR improvements for accented speech

Researchers have introduced PAREDA, a novel dataset designed to improve Automatic Speech Recognition (ASR) systems by capturing real-world speech variations. This dataset features discussions on Natural Language Processing (NLP) research papers among speakers with Australian, Indian-English, and Chinese English accents. PAREDA includes both spontaneous monologues and question-and-answer sessions, rich with technical jargon and conversational elements. Evaluations show that while state-of-the-art ASR models struggle in a zero-shot setting, fine-tuning on PAREDA significantly reduces word error rates, highlighting its value for developing more robust and inclusive ASR technologies for specialized applications. AI

影响 This dataset aims to improve ASR robustness for diverse accents, potentially enhancing accessibility and usability of speech technologies in global contexts.

排序理由 The cluster contains an academic paper introducing a new dataset for ASR research. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

• Natural Language Processing

• 论文
• 其他

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →