Sign In

PromptRad: Knowledge-Enhanced Multi-Label Prompt-Tuning for Low-Resource Radiology Report Labeling

์ž‘์„ฑ์ž
  • Haebom
์นดํ…Œ๊ณ ๋ฆฌ
Empty

์ €์ž

Ying-Jia Lin, Tzu-Chin Lo, Ping-Chien Li, Chi-Tung Cheng, Chien-Hung Liao, Hung-Yu Kao

๐Ÿ’ก ๊ฐœ์š”

๋ณธ ๋…ผ๋ฌธ์€ ์ €์ž์› ํ™˜๊ฒฝ์—์„œ ์˜์ƒ ํŒ๋… ๋ณด๊ณ ์„œ์˜ ๋‹ค์ค‘ ๋ ˆ์ด๋ธ” ๋ถ„๋ฅ˜๋ฅผ ์œ„ํ•œ ์ƒˆ๋กœ์šด ์ง€์‹ ๊ฐ•ํ™” ํ”„๋กฌํ”„ํŠธ ํŠœ๋‹ ๋ฐฉ๋ฒ•๋ก ์ธ PromptRad๋ฅผ ์ œ์•ˆํ•ฉ๋‹ˆ๋‹ค. PromptRad๋Š” UMLS ๋™์˜์–ด๋ฅผ ํ™œ์šฉํ•˜์—ฌ ๋ฒ”์ฃผ ํ‘œํ˜„์„ ํ’๋ถ€ํ•˜๊ฒŒ ํ•˜๊ณ , ๋งˆ์Šคํฌ ์–ธ์–ด ๋ชจ๋ธ๋ง ๋ฐฉ์‹์œผ๋กœ ๋‹ค์ค‘ ๋ ˆ์ด๋ธ” ๋ถ„๋ฅ˜ ๋ฌธ์ œ๋ฅผ ์žฌ๊ตฌ์„ฑํ•˜์—ฌ ์ถ”๊ฐ€ ๋ถ„๋ฅ˜ ๊ณ„์ธต ์—†์ด ์‚ฌ์ „ ํ›ˆ๋ จ๋œ ์–ธ์–ด ๋ชจ๋ธ์„ ๋ฏธ์„ธ ์กฐ์ •ํ•ฉ๋‹ˆ๋‹ค. ์‹คํ—˜ ๊ฒฐ๊ณผ, PromptRad๋Š” ๋งค์šฐ ์ ์€ ์–‘์˜ ๋ ˆ์ด๋ธ” ๋ฐ์ดํ„ฐ๋กœ๋„ ๊ธฐ์กด ๋ฐฉ๋ฒ•๋ก ๋ณด๋‹ค ์šฐ์ˆ˜ํ•œ ์„ฑ๋Šฅ์„ ๋ณด์ด๋ฉฐ, ๋ณต์žกํ•œ ๋ถ€์ • ํŒจํ„ด์„ ํšจ๊ณผ์ ์œผ๋กœ ํฌ์ฐฉํ•˜๋Š” ๊ฒƒ์œผ๋กœ ๋‚˜ํƒ€๋‚ฌ์Šต๋‹ˆ๋‹ค.

๐Ÿ”‘ ์‹œ์‚ฌ์  ๋ฐ ํ•œ๊ณ„

โ€ข
์ €์ž์› ํ™˜๊ฒฝ์˜ ์˜์ƒ ํŒ๋… ๋ณด๊ณ ์„œ ๋‹ค์ค‘ ๋ ˆ์ด๋ธ” ๋ถ„๋ฅ˜ ๋ฌธ์ œ ํ•ด๊ฒฐ์— ๊ธฐ์—ฌ
โ€ข
UMLS ๋™์˜์–ด ํ™œ์šฉ ๋ฐ ๋งˆ์Šคํฌ ์–ธ์–ด ๋ชจ๋ธ๋ง ๊ธฐ๋ฐ˜ ํ”„๋กฌํ”„ํŠธ ํŠœ๋‹์œผ๋กœ ๋ฐ์ดํ„ฐ ํšจ์œจ์„ฑ ๊ทน๋Œ€ํ™”
โ€ข
๋ณต์žกํ•œ ๋ถ€์ • ํŒจํ„ด ํฌ์ฐฉ ๋Šฅ๋ ฅ ํ–ฅ์ƒ์œผ๋กœ ์ž„์ƒ ํ™˜๊ฒฝ์—์„œ์˜ ํ™œ์šฉ ๊ฐ€๋Šฅ์„ฑ ์ œ์‹œ
โ€ข
ํ–ฅํ›„ ๋” ๋‹ค์–‘ํ•œ ์ž„์ƒ ๋ณด๊ณ ์„œ ์œ ํ˜• ๋ฐ ์–ธ์–ด ๋ชจ๋ธ๊ณผ์˜ ๊ฒฐํ•ฉ์— ๋Œ€ํ•œ ์—ฐ๊ตฌ ํ•„์š”
๐Ÿ‘
Made with Slashpage