Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)
ÇѱÛÁ¦¸ñ(Korean Title) |
¹®¼ ºÐ·ù¸¦ À§ÇÑ ½Å°æ¸Á ¸ðµ¨¿¡ ÀûÇÕÇÑ ÅØ½ºÆ® Àüó¸®¿Í ¿öµå ÀÓº£µùÀÇ Á¶ÇÕ |
¿µ¹®Á¦¸ñ(English Title) |
Combinations of Text Preprocessing and Word Embedding Suitable for Neural Network Models for Document Classification |
ÀúÀÚ(Author) |
±è¿µ¼ö
À̽¿ì
Yeongsu Kim
Seungwoo Lee
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 45 NO. 07 PP. 0690 ~ 0700 (2018. 07) |
Çѱ۳»¿ë (Korean Abstract) |
ÃÖ±Ù ¹®¼ ºÐ·ù¸¦ ÇØ°áÇϴµ¥ ½Å°æ¸Á ¸ðµ¨°ú ÇÔ²² ¿öµå ÀÓº£µùÀ» ÁÖ·Î »ç¿ëÇÑ´Ù. ¿¬±¸ÀÚµéÀº ¼º´ÉÀ» ³ôÀ̱â À§ÇØ »õ·Î¿î ½Å°æ¸Á ¸ðµ¨À» µðÀÚÀÎÇϰųª ¸ðµ¨ ÆÄ¶ó¹ÌÅ͸¦ ÃÖÀûÈÇϴµ¥ ½Ã°£À» ¸¹ÀÌ ÇÒ¾ÖÇÑ´Ù. ÇÏÁö¸¸, ¸¹Àº ¿¬±¸µéÀº Ưº°ÇÑ ÀÌÀ¯ ¾øÀÌ Æ¯Á¤ÇÑ ¿öµå ÀÓº£µù ¸ðµ¨À» »ç¿ëÇϰí Àü󸮿¡ ´ëÇÑ ÀÚ¼¼ÇÑ ¼³¸íÀ» ÇÏÁö ¾Ê´Â Á¡°ú °°ÀÌ Àüó¸®¿Í ¿öµå ÀÓº£µù¿¡ ´ëÇØ¼´Â ±×´ÙÁö ½Å°æÀ» ¾²°í ÀÖÁö ¾Ê´Ù. º» ¿¬±¸´Â ¼º´ÉÀ» Çâ»ó½ÃŰ´Â Ãß°¡ÀûÀÎ ¿ä¼Ò·Î ÀûÇÕÇÑ Àüó¸®¿Í ¿öµå ÀÓº£µù Á¶ÇÕÀ» ã´Â °ÍÀÓÀ» ¸»ÇϰíÀÚ ÇÑ´Ù. ½ÇÇèÀº À̵éÀÇ °¡´ÉÇÑ Á¶ÇÕµéÀ» ºñ±³½ÇÇèÇÏ´Â °Í°ú Á¦·Î/·£´ý ÆÐµù, ¹Ì¼¼ Á¶Á¤¿¡ ÀÇÇÑ ¿öµå ÀÓº£µù ÀçÇнÀ ¿©ºÎµµ °°ÀÌ ½ÇÇèÇÑ´Ù. ¶ÇÇÑ, »çÀü¿¡ ÇнÀÇÑ ¿öµå ÀÓº£µù ¸ðµ¨µé°ú ÇÔ²² Æò±Õ, ·£´ý, ÇнÀ µ¥ÀÌÅÍ·Î ÇнÀÇÑ ÀÓº£µùµéµµ °°ÀÌ »ç¿ëÇÑ´Ù. OOV(Out of Vocabulary)´Ü¾î Åë°è¸¦ ±âÁØÀ¸·Î ½ÇÇèÇÑ °á°ú·ÎºÎÅÍ À§¿Í °°Àº ½ÇÇèµéÀÇ Çʿ伺°ú Àüó¸®¿Í ¿öµå ÀÓº£µùÀÇ ÃÖÀûÀÇ Á¶ÇÕÀ» Á¦½ÃÇÑ´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
Neural networks with word embedding have recently used for document classification. Researchers concentrate on designing new architecture or optimizing model parameters to increase performance. However, most recent studies have overlooked text preprocessing and word embedding, in that the description of text preprocessing used is insufficient, and a certain pretrained word embedding model is mostly used without any plausible reasons. Our paper shows that finding a suitable combination of text preprocessing and word embedding can be one of the important factors required to enhance the performance. We conducted experiments on AG¡¯s News dataset to compare those possible combinations, and zero/random padding, and presence or absence of fine-tuning. We used pretrained word embedding models such as skip-gram, GloVe, and fastText. For diversity, we also use an average of multiple pretrained embeddings (Average), randomly initialized embedding (Random), task data-trained skip-gram (AGNews-Skip). In addition, we used three advanced neura networks for the sake of generality. Experimental results based on OOV (Out Of Vocabulary) word statistics suggest the necessity of those comparisons and a suitable combination of text preprocessing and word embedding.
|
Ű¿öµå(Keyword) |
¹®¼ ºÐ·ù
½Å°æ¸Á ¸ðµ¨
¿öµå ÀÓº£µù
ÅØ½ºÆ® Àüó¸®
document classification
neural network
word embedding
text preprocessing
|
¿ø¹® |
PDF ´Ù¿î·Îµå
|