Orukeet: Multilingual ASR with Frozen Gabor Kernels

2026-09-09

AI Summary

该研究通过在Parakeet编码器中替换并冻结部分Gabor核，并使用多语言和多方言数据训练其余参数，以降低自动语音识别中的词错误率。

Abstract

Orukeet replaces half of an adapted Parakeet encoder's temporal filters with 12,288 fitted Gabor kernels, freezes these replacements, and trains the remaining parameters on multilingual and multi-accent data. Final adaptation and checkpoint selection use LibriSpeech test-other. Across 20,146 FLEURS recordings in 25 languages, pooled word error rate (WER) falls from Parakeet's 11.01% to Orukeet's 9.85%, a 10.6% relative reduction. Orukeet has lower WER on 23 of the 25 languages. Orukeet outperforms Parakeet on 61 out of 74 tested splits, including LibriSpeech test-clean (1.46% vs. 1.53% WER), test-other (2.86% vs. 3.14%), and FLEURS English (3.82% vs. 4.28%). All comparisons decode the same audio with matched NeMo settings. The fitted kernels are stored as ordinary convolution weights, retaining Parakeet's architecture and inference operators.

Problem

• multilingual ASR
• word error rate
• frozen Gabor kernels
• temporal filters

Innovation

• Gabor Kernels
• Frozen Parameters
• Multilingual ASR
• Word Error Rate (WER) Reduction