@matematico314@social.linux.pizza2026-07-07 04:49:11
Nunca tinha parado para pensar que um subanel de um anel com unidade pode ter uma unidade diferente da unidade do anel mãe, pois pode ter um elemento que funcione como neutro multiplicativo apenas para elementos deste subanel. Por exemplo, considere o conjunto S das matrizes reais 2 x 2 em que todos os coeficientes, que não o superior esquerdo, sejam nulos. S é obviamente um subanel do anel das matrizes reais 2 x 2. Mas a sua unidade não é a matriz identidade, e sim o elemento de S cujo coef…
@arXiv_eessAS_bot@mastoxiv.page2026-05-13 07:47:08
Too Good to Be True: A Study on Modern Automatic Speech Recognition for the Evaluation of Speech Enhancement
Danilo de Oliveira, Tal Peer, Timo Gerkmann
https://arxiv.org/abs/2605.12107 https://arxiv.org/pdf/2605.12107 https://arxiv.org/html/2605.12107
arXiv:2605.12107v1 Announce Type: new
Abstract: Speech enhancement (SE) systems are typically evaluated using a variety of instrumental metrics. The use of automatic speech recognition (ASR) systems to evaluate SE performance is common in literature, usually in terms of word error rate (WER). However, WER scores depend heavily on the choice of ASR system and text normalization pipeline. In this paper, we investigate how modern ASR models correlate with human recognition of enhanced speech. A listening experiment reveals that modern ASR models with large-scale noisy training and embedded language models correlate more with human WER than simpler ones, with a transducer model providing the most reliable transcriptions. Nevertheless, we also show that these models' robustness to noise and use of context can be uninformative to an acoustics-focused evaluation of enhancement performance.
toXiv_bot_toot
@jamie@boothcomputing.social2026-04-21 18:21:09
