Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)
ÇѱÛÁ¦¸ñ(Korean Title) |
±íÀº ½Å°æ¸Á ±â¹Ý ´ë¿ë·® ÅؽºÆ® µ¥ÀÌÅÍ ºÐ·ù ±â¼ú |
¿µ¹®Á¦¸ñ(English Title) |
Large-Scale Text Classification with Deep Neural Networks |
ÀúÀÚ(Author) |
Á¶ÈÖ¿
±èÁøÈ
±è°æ¹Î
ÀåÁ¤È£
¾öÀçÈ«
À庴Ź
Hwiyeol Jo
Jin-Hwa Kim
Kyung-Min Kim
Jeong-Ho Chang
Jae-Hong Eom
Byoung-Tak Zhang
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 23 NO. 05 PP. 0322 ~ 0327 (2017. 05) |
Çѱ۳»¿ë (Korean Abstract) |
¹®¼ ºÐ·ù ¹®Á¦´Â ¿À·£ ±â°£ µ¿¾È ÀÚ¿¬¾î ó¸® ºÐ¾ß¿¡¼ ¿¬±¸µÇ¾î ¿Ô´Ù. ¿ì¸®´Â ±âÁ¸ ÄÁº¼·ç¼Ç ½Å°æ¸ÁÀ» ÀÌ¿ëÇß´ø ¿¬±¸¿¡¼ ³ª¾Æ°¡, ¼øȯ ½Å°æ¸Á¿¡ ±â¹ÝÀ» µÐ ¹®¼ ºÐ·ù¸¦ ¼öÇàÇÏ¿´°í ±× °á°ú¸¦ Á¾ÇÕÇÏ¿© Á¦½ÃÇÏ·Á ÇÑ´Ù. ÄÁº¼·ç¼Ç ½Å°æ¸ÁÀº ´ÜÃþ ÄÁº¼·ç¼Ç ½Å°æ¸ÁÀ» »ç¿ëÇßÀ¸¸ç, ¼øȯ ½Å°æ¸ÁÀº °¡Àå ¼º´ÉÀÌ ÁÁ´Ù°í ¾Ë·ÁÁ® ÀÖ´Â Àå±â-´Ü±â ±â¾ï ½Å°æ¸Á°ú ȸ·ÎÇü ¼øȯ À¯´ÖÀ» È°¿ëÇÏ¿´´Ù. ½ÇÇè °á°ú, ºÐ·ù Á¤È®µµ´Â Multinomial Naïve Bayesian Classifier < SVM < LSTM < CNN < GRUÀÇ ¼ø¼·Î ³ªÅ¸³µ´Ù. µû¶ó¼ ÅؽºÆ® ¹®¼ ºÐ·ù ¹®Á¦´Â ½ÃÄö½º¸¦ °í·ÁÇÏ´Â °Í º¸´Ù´Â ¹®¼ÀÇ feature¸¦ ÃßÃâÇÏ¿© ºÐ·ùÇÏ´Â ¹®Á¦¿¡ °¡±õ´Ù´Â °ÍÀ» È®ÀÎÇÒ ¼ö ÀÖ¾ú´Ù. ±×¸®°í GRU°¡ LSTMº¸´Ù ¹®¼ÀÇ feature ÃßÃâ¿¡ ´õ ÀûÇÕÇÏ´Ù´Â °ÍÀ» ¾Ë ¼ö ÀÖ¾úÀ¸¸ç ÀûÀýÇÑ feature¿Í ½ÃÄö½º Á¤º¸¸¦ ÇÔ²² È°¿ëÇÒ ¶§ °¡Àå ¼º´ÉÀÌ Àß ³ª¿Â´Ù´Â °ÍÀ» È®ÀÎÇÒ ¼ö ÀÖ¾ú´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
The classification problem in the field of Natural Language Processing has been studied for a long time. Continuing forward with our previous research, which classifies large-scale text using Convolutional Neural Networks (CNN), we implemented Recurrent Neural Networks (RNN), Long- Short Term Memory (LSTM) and Gated Recurrent Units (GRU). The experiment¡¯s result revealed that the performance of classification algorithms was Multinomial Naïve Bayesian Classifier < Support Vector Machine (SVM) < LSTM < CNN < GRU, in order. The result can be interpreted as follows: First, the result of CNN was better than LSTM. Therefore, the text classification problem might be related more to feature extraction problem than to natural language understanding problems. Second, judging from the results the GRU showed better performance in feature extraction than LSTM. Finally, the result that the GRU was better than CNN implies that text classification algorithms should consider feature extraction and sequential information. We presented the results of fine-tuning in deep neural networks to provide some intuition regard natural language processing to future researchers.
|
Å°¿öµå(Keyword) |
µö·¯´×
´ë¿ë·® ¹®¼ ºÐ·ù
ÀÚ¿¬¾î ó¸®
Àΰø½Å°æ¸Á
deep learning
large-scale text classification
natural language processing
artificial neural networks
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|