Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)
ÇѱÛÁ¦¸ñ(Korean Title) |
¹°Ã¼ °£ ¿¬°ü °ü°è ÇнÀÀ» ÅëÇÑ ¹®ÀåÀ¸·ÎºÎÅÍ Àå¸é »ý¼º |
¿µ¹®Á¦¸ñ(English Title) |
Scene Generation from a Sentence by Learning Object Relation |
ÀúÀÚ(Author) |
½Å¿ë¹Î
ÃÖ¼öÁ¤
¹Ú¼º¹è
¹Ú¼¼¿µ
Yongmin Shin
Su Jeong Choi
Seong-Bae Park
Seyoung Park
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 46 NO. 05 PP. 0431 ~ 0439 (2019. 05) |
Çѱ۳»¿ë (Korean Abstract) |
»ç¶÷°ú ±â°è°¡ ¼ÒÅëÇÒ ¶§ À§Ä¡ Á¤º¸´Â Áß¿äÇÑ ¿ªÇÒÀ» ÇÏÁö¸¸, ¶§¶§·Î ¹®Àå¿¡¼ »ý·«µÇ¾î ³ªÅ¸³´Ù. »ç¶÷Àº »ý·«µÈ Á¤º¸¸¦ ¹è°æ Áö½ÄÀ» ÅëÇؼ ¾Ë ¼ö ÀÖÁö¸¸, ±â°è´Â ±×·¸Áö ¾ÊÀº °æ¿ì°¡ ¸¹±â ¶§¹®¿¡ ¹®ÀåÀ¸·ÎºÎÅÍ Àå¸éÀ» »ý¼ºÇÑ´Ù¸é ¿Ã¹Ù¸¥ Àå¸éÀ» »ý¼ºÇÏÁö ¸ø ÇÑ´Ù. ÀÌ·¯ÇÑ ¹®Á¦¸¦ ÇØ°áÇϱâ À§ÇØ ±âÁ¸ÀÇ ¿¬±¸µéÀº ÁÖ¾îÁø ¹®Àå¿¡¼ ±¸ºÐ ºÐ¼® ¹æ½ÄÀ» ÅëÇØ ¸í½ÃÀûÀÎ À§Ä¡ °ü°è¸¦ ÃßÁ¤ÇÏ°í »çÀü È®·üµéÀ» ÀÌ¿ëÇÏ¿© »ý·«µÈ Á¤º¸¸¦ º¹¿øÇÏ¿© Àå¸éÀ» »ý¼ºÇÏ¿´À¸³ª, ÀÌ´Â ÇüÅ·ÐÀû »ý»ê¼º°ú ÀÚÀ¯·Î¿î ¾î¼ø µî ±¸¹® ºÐ¼®¹æ½ÄÀ» »ç¿ëÇϱ⿡ ¿©·¯ ¹®Á¦°¡ ÀÖ´Â Çѱ¹¾î ¹®Àå¿¡ Àû¿ëÇϱ⿡´Â ¸¹Àº ¾î·Á¿òÀÌ ÀÖ´Ù. µû¶ó¼ º» ³í¹®¿¡¼´Â Çѱ¹¾î ¹®ÀåÀ¸·ÎºÎÅÍ Àå¸éÀ» »ý¼ºÇÏ´Â ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. ¸ÕÀú ÁÖ¾îÁø ¹®Àå¿¡ ¸í½ÃÀûÀ¸·Î ³ªÅ¸³ ¹°Ã¼ ¹× À§Ä¡ °ü°è¸¦ ÃßÁ¤Çϱâ À§ÇØ RNN ±â¹ÝÀÇ ½Å°æ¸ÁÀ» ÀÌ¿ëÇÑ´Ù. ±× ´ÙÀ½ ¹°Ã¼ °£ ¿¬°ü °ü°è¿¡ ´ëÇÑ »çÀüÈ®·ü °ªµéÀ» ÇнÀÇÏ¿© ¹®Àå¿¡¼ »ý·«µÈ Á¤º¸¸¦ ÃßÁ¤ÇÑ´Ù. ÃßÁ¤µÈ ¹°Ã¼ ¹× À§Ä¡ Á¤º¸¸¦ »ç¿ëÇÏ¿© Àå¸é Æ®¸®¸¦ ±¸¼ºÇÏ°í À̸¦ ÅëÇØ Àå¸éÀ» »ý¼ºÇÑ´Ù. Àå¸é »ý¼º¿¡ ´ëÇÑ Æò°¡¸¦ À§Çؼ ¹®Àå¿¡ Á¸ÀçÇÏ´Â ¿¬°ü °ü°è¸¦ ´Ù·ç´Â ¸ðµ¨ÀÇ Á¤È®µµ¸¦ ÃøÁ¤ÇÏ°í »ý¼ºÇÑ Àå¸é¿¡ ´ëÇÑ Æò°¡¸¦ ¼öÇàÇÏ¿´´Ù. ±× °á°ú ±¸¹® ºÐ¼® ±â¹Ý ¹æ½Äº¸´Ù ¾à 25%ÀÇ ¼º´É Çâ»óÀÌ ÀÖ¾î Á¦¾ÈÇÑ ¹æ¹ýÀÌ È¿°úÀûÀÓÀ» Áõ¸íÇÏ¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
In communication between humans and machines, location information is crucial. However, it is sometimes omitted. While humans can infer omitted information, machines cannot. Thus, certain problems can occur when generating scenes from sentences. In order to solve this problem, previous studies have found an explicit relation in the sentence, then inferred an implicit relation by using prior probability. However, such methods are not suitable for Korean, as it has morphologically productivity. In this paper, we suggest a scene-generation method for Korean. Frist, we find an explicit relation by using an RNN-based artificial neural network. Then, to infer implicit information, we use the prior probability of relations. Finally, we prepare a scene tree with the obtained information, then generate a scene using that tree. In order to evaluate the scene generation, we measure the accuracy of the model dealing with the relationship and assign a human score to the generated scene. As a result, the method is proven to be effective with excellent performance and evaluation.
|
Å°¿öµå(Keyword) |
Àå¸é »ý¼º
À̹ÌÁö »ý¼º
À§Ä¡ °ü°è ÃßÁ¤
½ÉÃþ Àΰø ½Å°æ¸Á
scene generation
inferring location information
artificial neural network
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|