Seeking Standardized ASR Testing Methodologies and Optimization Guidelines

Hi everyone,
I’m currently exploring ways to test Automatic Speech Recognition (ASR) models, particularly focusing on real-time performance. I’m aware that there are existing benchmarks, but they mostly cater to the English language. I’m looking for standardized techniques or methodologies that can be used to compare different ASR models effectively, especially for other languages. Additionally, I would appreciate any guidelines or best practices for optimizing ASR models when integrated with other NLP functionalities and models. Does anyone have experience or resources to share on these topics?

Thanks in advance for your insights!