(评论)
(comments)

原始链接: https://news.ycombinator.com/item?id=43914738

aiola.ai开发的一个新的日语自动语音识别(ASR)模型“Jargonic”宣称达到了最先进(SOTA)的性能。《黑客新闻》的一篇文章报道了这一消息,并解释了SOTA和ASR的缩写。 一位评论者开玩笑地将“SOTA”误解为“空中峰会”(业余无线电活动)。另一位评论者质疑了SOTA的说法,指出缺乏与OpenAI的gpt-4o-transcribe的比较,认为需要与最新模型进行比较才能验证这一说法。另一位发帖者询问他们与其他现有模型相比做了哪些改进才能达到新的标准。

相关文章

原文
Hacker News new | past | comments | ask | show | jobs | submit login
Jargonic Sets New SOTA for Japanese ASR (aiola.ai)
19 points by four_fifths 1 day ago | hide | past | favorite | 4 comments










SOTA: not used in the article but probably State Of The Art

ASR: Automatic Speech Recognition, speech-to-text



And here I was, as a ham radio operator, excited to read something about Summits On The Air.

shuffles dejectedly back to shack



Why no comparition to gpt-4o-transcribe?

If you don't compare to latest model on the market, how can you claim it's SOTA?

According to OpenAI, gpt-4o-transcribe has much better performance than whisper-large-v2.

https://openai.com/index/introducing-our-next-generation-aud...



Are there any details on what they changed to improve over other existing models?






Consider applying for YC's Summer 2025 batch! Applications are open till May 13


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact



Search:
联系我们 contact @ memedata.com