Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)
ÇѱÛÁ¦¸ñ(Korean Title) |
´ÙÁß Å¬·¡½º ¸â¹ö½± 󸮸¦ À§ÇÑ Bi-LSTM ±â¹Ý Áö½Ä ±×·¡ÇÁ ¿Ï¼º ±â¹ý |
¿µ¹®Á¦¸ñ(English Title) |
Approach for Managing Multiple Class Membership in Knowledge Graph Completion Using Bi-LSTM |
ÀúÀÚ(Author) |
±èÅ¿µ
Á¶¼º¹è
Tae-Young Kim
Sung-Bae Cho
³ëÀç½Â
¹ÙÆ®¼¿·½
ÀÌ¿Ï°ï
¹Ú¿µÅÃ
Jae-Seung Roh
Batselem Jagvaral
Wan-Gon Lee
Young-Tack Park
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 47 NO. 06 PP. 0559 ~ 0567 (2020. 06) |
Çѱ۳»¿ë (Korean Abstract) |
½Ç¼¼°èÀÇ Áö½ÄÀ» ±¸Á¶ÈµÈ ¹æ½ÄÀ¸·Î Ç¥ÇöÇÑ Áö½Ä ±×·¡ÇÁ´Â À¥ °Ë»ö, Ãßõ ½Ã½ºÅÛ°ú °°ÀÌ ´Ù¾çÇÑ ºÐ¾ß¿¡¼ È°¿ëµÇ°í ÀÖÁö¸¸, ¿£Æ¼Æ¼ ¶Ç´Â ¿£Æ¼Æ¼ »çÀÌÀÇ ¸µÅ©°¡ ´©¶ôµÇ´Â ¹®Á¦°¡ Á¸ÀçÇÑ´Ù. ÀÌ·¯ÇÑ ¹®Á¦ÇØ°áÀ» À§ÇØ ÀÓº£µù ±â¹ýÀ» »ç¿ëÇϰųª µö·¯´×À» È°¿ëÇÑ ´Ù¾çÇÑ ¿¬±¸µéÀÌ ÁøÇàµÇ¾úÀ¸¸ç, ƯÈ÷ CNN°ú Bidirectional-LSTMÀ» °áÇÕÇÑ ÃֽŠ¿¬±¸°¡ ±âÁ¸ ¿¬±¸µé°ú ºñ±³ÇÏ¿© ³ôÀº ¼º´ÉÀ» ³ªÅ¸³Â´Ù. ±×·¯³ª ÇϳªÀÇ ¿£Æ¼Æ¼¿¡ ´ëÇÏ¿© ¿©·¯ °³ÀÇ Å¬·¡½º ŸÀÔÀÌ Á¤ÀÇµÈ °æ¿ì ÇнÀ µ¥ÀÌÅÍÀÇ ¾çÀÌ ±âÇϱ޼öÀûÀ¸·Î Áõ´ëµÇ¾î ÇнÀ½Ã°£ÀÌ Áõ°¡ÇÏ´Â ¹®Á¦¿Í ¿£Æ¼Æ¼ÀÇ Å¬·¡½º ŸÀÔ Á¤º¸°¡ Á¤ÀǵÇÁö ¾ÊÀ¸¸é ÇнÀ µ¥ÀÌÅÍ »ý¼ºÀÌ ºÒ°¡´ÉÇÏ´Ù´Â ÇÑ°èÁ¡ÀÌ Á¸ÀçÇÑ´Ù. µû¶ó¼ º» ³í¹®¿¡¼´Â ¿£Æ¼Æ¼ÀÇ Å¬·¡½º ŸÀÔ ¼ö¿¡ »ó°ü¾øÀÌ ÇнÀ µ¥ÀÌÅÍ »ý¼º°ú ¸ðµ¨¿¡¼ ÇнÀ ¹× Ãß·ÐÀÌ °¡´ÉÇϵµ·Ï ¹Ì¸® ÇнÀµÈ Áö½Ä ±×·¡ÇÁ ÀÓº£µù º¤Å͸¦ »ç¿ëÇÏ´Â ¹æ¹ý°ú vector addition °³³äÀ» È°¿ëÇÑ ´ÙÁß Å¬·¡½º ¸â¹ö½± ó¸® ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. º» ³í¹®¿¡¼ Á¦¾ÈÇÏ´Â ¹æ¹ýÀÇ ¼º´ÉÀ» Æò°¡Çϱâ À§ÇØ µ¥ÀÌÅͼ NELL-995 ¿Í FB15K-237À» ´ë»óÀ¸·Î ±âÁ¸ Áö½Ä ¿Ï¼º ¿¬±¸µé°ú ºñ±³ ½ÇÇèÀ» ÁøÇàÇÏ¿´À¸¸ç MAPÀÌ 1.6%p, MRRÀÌ 1.5%p ´õ ³ôÀº ¼º´ÉÀ» º¸¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
Knowledge graphs that represent real world information in a structured way are widely used in areas, such as Web browsing and recommendation systems. But there is a problem of missing links between entities in knowledge graphs. To resolve this issue, various studies using embedding techniques or deep learning have been proposed. Especially, the recent study combining CNN and Bidirectional-LSTM has shown high performance compared to previous studies. However, in the previous study, if multiple class types are defined for single entity, the amount of training data exponentially increases with the training time. Also, if class type information for an entity is not defined, training data for that entity cannot be generated. Thus, to enable the generation of training data for such entities and manage multiple class membership in knowledge graph completion, we propose two approaches using pre-trained embedding vectors of knowledge graph and the concept of vector addition. To evaluate the performance of the methods proposed in this paper, we conducted comparative experiments with the existing knowledge completion studies on NELL-995 and FB15K-237 datasets, and obtained MAP 1.6%p and MRR 1.5%p higher than that of the previous studies.
|
Å°¿öµå(Keyword) |
¶óÀÌÇÁ·Î±×
ºòµ¥ÀÌÅÍ
ÆÐÅÏ ¸¶ÀÌ´×
½Ã¸Çƽ ³×Æ®¿öÅ©
ÆäÆ®¸®³Ý
life log
big data
pattern mining
semantic network
petri-net
Áö½Ä ±×·¡ÇÁ
Áö½Ä ¿Ï¼º
¸µÅ© ¿¹Ãø
ÀÓº£µù
knowledge graph
knowledge completion
µö·¯´×
link prediction
embedding
deep learning
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|