Edit model card

Model Card for Model ID

์ด ๋ชจ๋ธ์€ KM-BERT์˜ ๋ชจ๋ธ์„ ๋‹ค์šด ๋ฐ›์•„, ์ž„์˜๋กœ ํ—ˆ๊น…ํŽ˜์ด์Šค ๋ชจ๋ธ์˜ ํ˜•ํƒœ๋กœ ์‚ฌ์šฉํ•˜๊ธฐ ์ข‹๊ฒŒ ๋ณ€ํ™˜ํ•œ ๋ชจ๋ธ ์ž…๋‹ˆ๋‹ค.

๋ชจ๋ธ์€ ๊ทธ๋Œ€๋กœ์ด๊ณ , ํ† ํฌ๋‚˜์ด์ €๋Š” ๋…ผ๋ฌธ์— ๋‚˜์™€ ์žˆ๋Š”๋Œ€๋กœ, snunlp/KR-BERT-char16424 ๊ฒƒ์„ ๊ฐ€์ ธ์™”์Šต๋‹ˆ๋‹ค

KM-BERT๋ฅผ ์‚ฌ์šฉํ•ด๋ณผ ์ˆ˜ ์žˆ๊ฒŒ ๊ณต์œ ํ•ด์ฃผ์‹  ์ €์ž๋ถ„๊ป˜ ๊ฐ์‚ฌ๋“œ๋ฆฝ๋‹ˆ๋‹ค.

๋ฌธ์ œ ๋ฐœ์ƒ ์‹œ ์ˆ˜์ •ํ•˜๊ฑฐ๋‚˜, ๋˜๋Š” ์ €์ž๋‹˜์˜ ์š”์ฒญ์‹œ ํ—ˆ๋ธŒ์—์„œ ๋‚ด๋ฆด ์ˆ˜๋„ ์žˆ์Šต๋‹ˆ๋‹ค.

Model Details

bing ํ™œ์šฉ ์š”์•ฝ:

์ด ๋ฌธ์„œ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๋‚ด์šฉ์„ ๋‹ค๋ฃน๋‹ˆ๋‹ค:

ํ•œ๊ตญ์–ด ์˜๋ฃŒ ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ๋ฅผ ์œ„ํ•œ ์‚ฌ์ „ ํ›ˆ๋ จ๋œ BERT ๋ชจ๋ธ.
์ด ๋ชจ๋ธ์€ ์˜๋ฃŒ ๋ถ„์•ผ์—์„œ ์‚ฌ์šฉ๋˜๋Š” ํ•œ๊ตญ์–ด ํ…์ŠคํŠธ๋ฅผ ๋ถ„์„ํ•˜๊ธฐ ์œ„ํ•ด KR-BERT ๋ชจ๋ธ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ๊ตญ์–ด ์˜๋ฃŒ ์ฝ”ํผ์Šค๋กœ ์‚ฌ์ „ ํ›ˆ๋ จํ•œ KM-BERT๋ผ๊ณ  ํ•ฉ๋‹ˆ๋‹ค.
์ด ๋ชจ๋ธ์€ ์˜๋ฃŒ ๊ต๊ณผ์„œ, ๊ฑด๊ฐ• ์ •๋ณด ๋‰ด์Šค, ์˜๋ฃŒ ์—ฐ๊ตฌ ๋…ผ๋ฌธ ๋“ฑ ์„ธ ๊ฐ€์ง€ ์œ ํ˜•์˜ ๋ฌธ์„œ๋กœ ๊ตฌ์„ฑ๋œ ์•ฝ 1์–ต 1์ฒœ 6๋ฐฑ๋งŒ ๋‹จ์–ด์˜ ํ•œ๊ตญ์–ด ์˜๋ฃŒ ์ฝ”ํผ์Šค๋ฅผ ์‚ฌ์šฉํ–ˆ์Šต๋‹ˆ๋‹ค.

๋‚ด๋ถ€ ํ‰๊ฐ€์™€ ์™ธ๋ถ€ ํ‰๊ฐ€.
์ด ๋ชจ๋ธ์€ MLM๊ณผ NSP๋ผ๋Š” ๋‘ ๊ฐ€์ง€ ๋น„์ง€๋„ ์‚ฌ์ „ ํ›ˆ๋ จ ๊ณผ์ œ๋ฅผ ์ˆ˜ํ–‰ํ•˜๊ณ , ์ƒˆ๋กœ์šด ํ•œ๊ตญ์–ด ์˜๋ฃŒ ์ฝ”ํผ์Šค์™€ MedSTS ๋ฐ์ดํ„ฐ์…‹์„ ์ด์šฉํ•˜์—ฌ ์–ธ์–ด ์ดํ•ด ๋Šฅ๋ ฅ์„ ๋‚ด๋ถ€ ํ‰๊ฐ€ํ–ˆ์Šต๋‹ˆ๋‹ค.
๋˜ํ•œ, ํ•œ๊ตญ์–ด ์˜๋ฃŒ ์˜๋ฏธ๋ก ์  ํ…์ŠคํŠธ ์œ ์‚ฌ๋„(MedSTS) ๋ฐ์ดํ„ฐ์…‹๊ณผ ํ•œ๊ตญ์–ด ์˜๋ฃŒ ๊ฐœ์ฒด ์ธ์‹(NER) ๋ฐ์ดํ„ฐ์…‹์„ ์ด์šฉํ•˜์—ฌ ํ•˜๋ฅ˜ ๊ณผ์ œ์— ๋Œ€ํ•œ ์„ฑ๋Šฅ์„ ์™ธ๋ถ€ ํ‰๊ฐ€ํ–ˆ์Šต๋‹ˆ๋‹ค.
๊ฒฐ๊ณผ์ ์œผ๋กœ, KM-BERT๋Š” ๊ธฐ์กด์˜ ์–ธ์–ด ๋ชจ๋ธ๋“ค๋ณด๋‹ค ์šฐ์ˆ˜ํ•œ ์„ฑ๋Šฅ์„ ๋ณด์˜€์Šต๋‹ˆ๋‹ค.

์–ธ์–ด๋ณ„ ๋ฐ ๋ถ„์•ผ๋ณ„ ๊ด€์ .
M-BERT๋Š” ๋‹ค์–‘ํ•œ ์–ธ์–ด์— ๋Œ€ํ•œ NLP ๊ณผ์ œ์— ๋†’์€ ์ „์ด์„ฑ์„ ๋ณด์ด์ง€๋งŒ, ์–ธ์–ด๋ณ„ BERT ๋ชจ๋ธ๋“ค์ด M-BERT๋ณด๋‹ค ๋” ๋‚˜์€ ์„ฑ๋Šฅ์„ ๋ณด์ธ๋‹ค๋Š” ๊ฒƒ์ด ์ฆ๋ช…๋˜์—ˆ์Šต๋‹ˆ๋‹ค.

  1. ํ•œ๊ตญ์–ด๋Š” ์˜์–ด์™€ ๊ฐ™์€ ์ธ๋„์œ ๋Ÿฝ์–ด์™€๋Š” ๋‹ค๋ฅธ ๋‹จ์–ด ์ˆœ์„œ๋ฅผ ๊ฐ€์ง€๊ณ  ์žˆ๊ธฐ ๋•Œ๋ฌธ์—, ํ•œ๊ตญ์–ด์— ํŠนํ™”๋œ BERT ๋ชจ๋ธ์ด ํ•„์š”ํ•ฉ๋‹ˆ๋‹ค.
  2. ๋˜ํ•œ, SciBERT๋‚˜ LEGAL-BERT์™€ ๊ฐ™์€ ๋ถ„์•ผ๋ณ„ BERT ์ž„๋ฒ ๋”ฉ์ด BERT๋ณด๋‹ค ์ „๋ฌธ ์šฉ์–ด์™€ ์šฉ๋ฒ•์— ๋Œ€ํ•œ ์ดํ•ด๋ ฅ์„ ํ–ฅ์ƒ์‹œํ‚จ๋‹ค๋Š” ๊ฒƒ์ด ๋ฐํ˜€์กŒ์Šต๋‹ˆ๋‹ค.
  3. ์ด๋Ÿฌํ•œ ์—ฐ๊ตฌ ๊ฒฐ๊ณผ๋“ค์€ ํ•œ๊ตญ์–ด ์˜๋ฃŒ ๋ถ„์•ผ์— ์ ํ•ฉํ•œ BERT ๋ชจ๋ธ ๊ฐœ๋ฐœ์˜ ํ•„์š”์„ฑ๊ณผ ํšจ๊ณผ์„ฑ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.

Model Sources [optional]

Citation [optional]

@article{KMBERT,
title={KM-BERT: A Pre-trained BERT for Korean Medical Natural Language Processing},
author={TBD},
year={TBD},
journal={TBD},
volume={TBD}
}

Downloads last month
577
Safetensors
Model size
98.7M params
Tensor type
I64
ยท
F32
ยท
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for madatnlp/km-bert

Finetunes
1 model