• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ³í¹®Áö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ³í¹®Áö > Çѱ¹Á¤º¸°úÇÐȸ ³í¹®Áö > Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)

Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) ÀÇ·á Á¶¾ðÀ» À§ÇÑ Áú¹® Àǵµ ÀνÄ: ÇнÀ µ¥ÀÌÅÍ ±¸Ãà ¹× Àǵµ ºÐ·ù
¿µ¹®Á¦¸ñ(English Title) Query Intent Detection for Medical Advice: Training Data Construction and Intent Classification
ÀúÀÚ(Author) ÀÌÅÂÈÆ   ±è¿µ¹Î   Á¤ÀºÁö   ³ª¼±¿Á   Tae-Hoon Lee   Young-Min Kim   Eunji Jeong   Seon-Ok Na  
¿ø¹®¼ö·Ïó(Citation) VOL 48 NO. 08 PP. 0878 ~ 0884 (2021. 08)
Çѱ۳»¿ë
(Korean Abstract)
´ëºÎºÐÀÇ °ú¾÷ ÁöÇâ ´ëÈ­ ½Ã½ºÅÛ¿¡¼­´Â Àǵµ Àνİú °³Ã¼¸í ÀνÄÀÌ ¼±ÇàµÇ¾î¾ß ÇÑ´Ù. º» ¿¬±¸¿¡¼­´Â ÀÇ·á Á¶¾ðÀ̶ó´Â ½Å±Ô ºÐ¾ß¿¡ ´ëÇÑ ´ëÈ­ ½Ã½ºÅÛ ±¸ÃàÀ» À§ÇØ »ç¿ëÀÚ Áú¹®ÀÇ Àǵµ¸¦ ÀνÄÇÏ´Â ¹®Á¦¸¦ ´Ù·é´Ù. ÃÖÁ¾ ¸ñÀû¿¡ ÇØ´çÇÏ´Â ÀÇ·á Á¶¾ðÀ» À§ÇØ ÇÊ¿äÇÑ Àǵµ Ä«Å×°í¸®¸¦ Á¤ÀÇÇÏ´Â °Í¿¡¼­ºÎÅÍ ÇнÀµ¥ÀÌÅÍ ¼öÁý ¹× ±¸Ãà, ·¹ÀÌºí¸µÀ» À§ÇÑ °¡À̵å¶óÀÎÀ» »ó¼úÇÑ´Ù. Áú¹® Àǵµ ÀνÄÀ» À§ÇØ BERT ±â¹ÝÀÇ ºÐ·ù ¸ðµ¨À» »ç¿ëÇßÀ¸¸ç Çѱ¹¾î 󸮸¦ À§ÇØ º¯ÇüµÈ KorBERTµµ Àû¿ëÇÑ´Ù. µö·¯´× ±â¹ÝÀÇ ¸ðµ¨ÀÌ º» ¿¬±¸¿¡¼­ ±¸ÃàÇÑ Áß±Ô¸ðÀÇ ÇнÀ µ¥ÀÌÅÍ¿¡¼­µµ ÁÁÀº ¼º´ÉÀ» º¸ÀÌ´Â °ÍÀ» °ËÁõÇϱâ À§ÇØ ÀϹÝÀûÀ¸·Î ¸¹ÀÌ ¾²ÀÌ´Â SVMµµ ºñ±³ ¸ðµ¨·Î È°¿ëÇÏ¿´´Ù. ½ÇÇè °á°ú 8°³ÀÇ Àǵµ Ä«Å×°í¸®¿¡ ´ëÇÑ f1 Á¡¼ö°¡ SVM, BERT, KorBERT¿¡¼­ °¢±â 69%, 78%, 84% ¿´À¸¸ç ÇâÈÄ µ¥ÀÌÅÍ º¸°­À» ÅëÇØ ÃÖÁ¾ ¼º´ÉÀ» ³ôÀÏ ¿¹Á¤ÀÌ´Ù.
¿µ¹®³»¿ë
(English Abstract)
In most task-oriented dialogue systems, intent detection and named entity recognition need to precede. This paper deals with the query intent detection to construct a dialogue system for medical advice. We start from the appropriate intent categories for the final goal. We also describe in detail the data collection, training data construction, and the guidelines for the manual annotation. BERT-based classification model has been used for query intent detection. KorBERT, a Korean version of BERT has been also tested for detection. To verify that the DNN-based models outperform the traditional machine learning methods even for a mid-sized dataset, we also tested SVM, which produces a good result in general for such dataset. The F1 scores of SVM, BERT, and KorBERT are 69%, 78%, and 84% respectively. For future work, we will try to increase intent detection performance through dataset improvement.
Å°¿öµå(Keyword) Áú¹® Àǵµ ÀνĠ  ÀÇ·á Á¶¾ð   ·¹ÀÌºí¸µ °¡À̵å¶óÀΠ  °ú¾÷ ÁöÇâ ´ëÈ­ ½Ã½ºÅÛ   query intent detection   medical advice   guidelines for labeling   task-oriented dialogue system  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå