• 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¸Ã³¸®ÇÐȸ ³í¹®Áö > Á¤º¸Ã³¸®ÇÐȸ ³í¹®Áö ¼ÒÇÁÆ®¿þ¾î ¹× µ¥ÀÌÅÍ °øÇÐ

Á¤º¸Ã³¸®ÇÐȸ ³í¹®Áö ¼ÒÇÁÆ®¿þ¾î ¹× µ¥ÀÌÅÍ °øÇÐ

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) Çѱ¹¾î ÇнÀ ¸ðµ¨º° Çѱ¹¾î ¾²±â ´ä¾ÈÁö Á¡¼ö ±¸°£ ¿¹Ãø ¼º´É ºñ±³
¿µ¹®Á¦¸ñ(English Title) Comparison of Korean Classification Models¡¯ Korean Essay Score Range Prediction Performance
ÀúÀÚ(Author) Á¶Èñ·Ã   ÀÓÇö¿­   ÀÌÀ¯¹Ì   Â÷ÁØ¿ì   Heeryon Cho   Hyeonyeol Im   Yumi Yi   Junwoo Cha  
¿ø¹®¼ö·Ïó(Citation) VOL 11 NO. 03 PP. 0133 ~ 0140 (2022. 03)
Çѱ۳»¿ë
(Korean Abstract)
¿ì¸®´Â À¯ÇлýÀÌ ÀÛ¼ºÇÑ Çѱ¹¾î ¾²±â ´ä¾ÈÁöÀÇ Á¡¼ö ±¸°£À» ¿¹ÃøÇÏ´Â ¹®Á¦¿¡¼­ ¼¼ °³ÀÇ µö·¯´× ±â¹Ý Çѱ¹¾î ¾ð¾î¸ðµ¨ÀÇ ¿¹Ãø ¼º´ÉÀ» Á¶»çÇÑ´Ù. À̸¦ À§ÇØ ÃÑ 304ÆíÀÇ ´ä¾ÈÁö·Î ±¸¼ºµÈ ½ÇÇè µ¥ÀÌÅÍ ¼¼Æ®¸¦ ±¸ÃàÇÏ¿´´Âµ¥, ´ä¾ÈÁöÀÇ ÁÖÁ¦´Â Á÷¾÷ ¼±ÅÃÀÇ ±âÁØ(¡®Á÷¾÷¡¯), ÇູÇÑ »îÀÇ Á¶°Ç(¡®Çູ¡¯), µ·°ú Çູ(¡®°æÁ¦¡¯), ¼º°øÀÇ Á¤ÀÇ(¡®¼º°ø¡¯)·Î ´Ù¾çÇÏ´Ù. ÀÌµé ´ä¾ÈÁö´Â ³× °³ÀÇ Á¡¼ö ±¸°£À¸·Î ±¸ºÐµÇ¾î Æò¾î ·¹À̺í(A, B, C, D)ÀÌ ¸Å°ÜÁ³°í, ÃÑ 11°ÇÀÇ Á¡¼ö ±¸°£ ¿¹Ãø ½ÇÇèÀÌ ½ÃÇàµÇ¾ú´Ù. ±¸Ã¼ÀûÀ¸·Î´Â 5°³ÀÇ ¡®Á÷¾÷¡¯ ´ä¾ÈÁö Á¡¼ö ±¸°£(Æò¾î) ¿¹Ãø ½ÇÇè, 5°³ÀÇ ¡®Çູ¡¯ ´ä¾ÈÁö Á¡¼ö ±¸°£ ¿¹Ãø ½ÇÇè, 1°³ÀÇ È¥ÇÕ ´ä¾ÈÁö Á¡¼ö ±¸°£ ¿¹Ãø ½ÇÇèÀÌ ½ÃÇàµÇ¾ú´Ù. ÀÌµé ½ÇÇè¿¡¼­ ¼¼ °³ÀÇ µö·¯´× ±â¹Ý Çѱ¹¾î ¾ð¾î¸ðµ¨(KoBERT, KcBERT, KR-BERT)ÀÌ ´Ù¾çÇÑ ÈÆ·Ã µ¥ÀÌÅÍ·Î ¹Ì¼¼Á¶Á¤µÇ¾ú´Ù. ¶Ç µÎ °³ÀÇ ÀüÅëÀûÀÎ È®·üÀû ±â°èÇнÀ ºÐ·ù±â(³ªÀÌºê º£ÀÌÁî¿Í ·ÎÁö½ºÆ½ ȸ±Í)µµ ±× ¼º´ÉÀÌ ºÐ¼®µÇ¾ú´Ù. ½ÇÇè °á°ú µö·¯´× ±â¹Ý Çѱ¹¾î ¾ð¾î¸ðµ¨ÀÌ ÀüÅëÀûÀÎ ±â°èÇнÀ ºÐ·ù±âº¸´Ù ¿ì¼öÇÑ ¼º´ÉÀ» º¸¿´À¸¸ç, ƯÈ÷ KR-BERT´Â Àü¹ÝÀûÀÎ Æò±Õ ¿¹Ãø Á¤È®µµ°¡ 55.83%·Î °¡Àå ¿ì¼öÇÑ ¼º´ÉÀ» º¸¿´´Ù. ±× ´ÙÀ½Àº KcBERT(55.77%)¿´°í KoBERT(54.91%)°¡ µÚ¸¦ À̾ú´Ù. ³ªÀÌºê º£ÀÌÁî¿Í ·ÎÁö½ºÆ½ ȸ±Í ºÐ·ù±âÀÇ ¼º´ÉÀº °¢°¢ 52.52%¿Í 50.28%¿´´Ù. ÇнÀµÈ ºÐ·ù±â ¸ðµÎ ÈÆ·Ã µ¥ÀÌÅÍÀÇ ºÎÁ·°ú µ¥ÀÌÅÍ ºÐÆ÷ÀÇ ºÒ±ÕÇü ¶§¹®¿¡ ¿¹Ãø ¼º´ÉÀÌ º°·Î ³ôÁö ¾Ê¾Ò°í, ºÐ·ù±âÀÇ ¾îÈÖ°¡ ±Û¾²±â ´ä¾ÈÁöÀÇ ¿À·ù¸¦ Á¦´ë·Î Æ÷ÂøÇÏÁö ¸øÇÏ´Â ÇÑ°è°¡ ÀÖ¾ú´Ù. ÀÌ µÎ °¡Áö ÇѰ踦 ±Øº¹ÇÏ¸é ºÐ·ù±âÀÇ ¼º´ÉÀÌ Çâ»óµÉ °ÍÀ¸·Î º¸ÀδÙ.
¿µ¹®³»¿ë
(English Abstract)
We investigate the performance of deep learning-based Korean language models on a task of predicting the score range of Korean essays written by foreign students. We construct a data set containing a total of 304 essays, which include essays discussing the criteria for choosing a job (¡®job¡¯), conditions of a happy life (¡®happ¡¯), relationship between money and happiness (¡®econ¡¯), and definition of success (¡®succ¡¯). These essays were labeled according to four letter grades (A, B, C, and D), and a total of eleven essay score range prediction experiments were conducted (i.e., five for predicting the score range of ¡®job¡¯ essays, five for predicting the score range of ¡®happiness¡¯ essays, and one for predicting the score range of mixed topic essays). Three deep learning-based Korean language models, KoBERT, KcBERT, and KR-BERT, were fine-tuned using various training data. Moreover, two traditional probabilistic machine learning classifiers, naive Bayes and logistic regression, were also evaluated. Experiment results show that deep learning-based Korean language models performed better than the two traditional classifiers, with KR-BERT performing the best with 55.83% overall average prediction accuracy. A close second was KcBERT (55.77%) followed by KoBERT (54.91%). The performances of naïve Bayes and logistic regression classifiers were 52.52% and 50.28% respectively. Due to the scarcity of training data and the imbalance in class distribution, the overall prediction performance was not high for all classifiers. Moreover, the classifiers¡¯ vocabulary did not explicitly capture the error features that were helpful in correctly grading the Korean essay. By overcoming these two limitations, we expect the score range prediction performance to improve.
Å°¿öµå(Keyword) Çѱ¹¾î ½ÉÃþÇнÀ ¾ð¾î¸ðµ¨   KoBERT   KcBERT   KR-BERT   ¹®¼­ ºÐ·ù   Deep Learning-Based Korean Language Model   KoBERT   KcBERT   KR-BERT   Document Classification  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå