Contact us

Tag: usm

News
Google Researcher's In-Depth Analysis on End-to-End Speech Recognition, Part 1: Overview & Modeling
Tutorials

Google ์—ฐ๊ตฌ์›์˜ End-to-End ์Œ์„ฑ ์ธ์‹์— ๋Œ€ํ•œ ์‹ฌ์ธต ๋ถ„์„, Part 1: ๊ฐœ์š” & ๋ชจ๋ธ๋ง

๋”ฅ๋Ÿฌ๋‹์˜ 10๋…„์€ ASR ๋‹จ์–ด ์˜ค๋ฅ˜์œจ์„ 50% ์ด์ƒ ์ƒ๋Œ€์ ์œผ๋กœ ๊ฐ์†Œ์‹œ์ผฐ๊ณ , ๊ทธ ๊ณผ์ •์—์„œ ๊ณ ์ „์ ์ธ HMM ํŒŒ์ดํ”„๋ผ์ธ์„ ์œ ๋ฌผ์ฒ˜๋Ÿผ ๋ณด์ด๊ฒŒ ๋งŒ๋“ค์—ˆ์Šต๋‹ˆ๋‹ค.

6์›” 7, 2024ยท 1 min read
Google's Universal Speech Model for 100+ languages beats OpenAI's Whisper Model
Tutorials

Google์˜ 100๊ฐœ ์ด์ƒ ์–ธ์–ด๋ฅผ ์ง€์›ํ•˜๋Š” Universal Speech Model์ด OpenAI์˜ Whisper Model์„ ๋Šฅ๊ฐ€ํ•ฉ๋‹ˆ๋‹ค

Google์˜ Universal Speech Model์€ ์กฐ์šฉํžˆ ๋†€๋ผ์šด ์„ฑ๊ณผ๋ฅผ ๋‹ฌ์„ฑํ–ˆ์Šต๋‹ˆ๋‹ค: ๋ ˆ์ด๋ธ”์ด ์ง€์ •๋œ ํ›ˆ๋ จ ๋ฐ์ดํ„ฐ์˜ ์•ฝ 1/7์„ ์‚ฌ์šฉํ•˜๋ฉด์„œ 100๊ฐœ ์ด์ƒ์˜ ์–ธ์–ด์—์„œ OpenAI์˜ Whisper์™€ ๋™๋“ฑํ•˜๊ฑฐ๋‚˜ ๋Šฅ๊ฐ€ํ•˜๋Š” ASR ์„ฑ๋Šฅ์„ ๋‹ฌ์„ฑํ–ˆ์Šต๋‹ˆ๋‹ค.

5์›” 6, 2024ยท 1 min read