µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
ºÎÈ£Çü ±×·¡ÇÁ ¿ä¾à ±â¹ý |
¿µ¹®Á¦¸ñ(English Title) |
Summarization of Signed Graphs |
ÀúÀÚ(Author) |
񊯈
±èÄ¡ÈÆ
¿ì¹Î½Ä
ÀÌ¿ìÁø
½Åµ¿¿ì
ÀÌ»óÁØ
San Kim
Chihun Kim
Minsik Woo
Woojin Lee
Dongwoo Shin
Sangjun Lee
Á¶Çö¼ö
ÁÖÇöÁø
¾ÈÁ¾Ã¶
Áø¼Ò¿¬
½ÅÀ¯°æ
½Å±âÁ¤
Hyeonsoo Jo
Hyunjin Choo
Jong-Chul Ahn
Soyeon Jin
Yukyung Shin
Kijung Shin
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 38 NO. 03 PP. 0003 ~ 0015 (2022. 12) |
Çѱ۳»¿ë (Korean Abstract) |
¼Ò¼È ³×Æ®¿öÅ© ¼ºñ½º, ÀüÀÚ »ó°Å·¡ µî ´Ù¾çÇÑ À¥ ¼ºñ½ºµéÀÇ »ç¿ë·®ÀÌ Æø¹ßÀûÀ¸·Î Áõ°¡ÇÏ°í ÀÖÀ¸¸ç, ÇØ´ç Á¤º¸µéÀ» Ç¥ÇöÇϱâ À§ÇÑ ±×·¡ÇÁÀÇ ±Ô¸ð ¿ª½Ã ±âÇϱ޼öÀûÀ¸·Î Áõ°¡ÇÏ°í ÀÖ´Ù. ÀÌ·¯ÇÑ ´ë¿ë·® ±×·¡ÇÁ¸¦ È¿À²ÀûÀ¸·Î È°¿ëÇϱâ À§ÇØ Á¤º¸ ¼Õ½ÇÀ» ÃÖ¼ÒÈÇÏ¸é¼ °£°áÇÑ ¿ä¾à ±×·¡ÇÁ¸¦ ¾ò¾î³»´Â ±×·¡ÇÁ ¿ä¾à ¹æ¹ýÀÌ Á¦¾ÈµÇ¾ú´Ù. ÇÏÁö¸¸ ±âÁ¸ ±×·¡ÇÁ ¿ä¾à ¹æ¹ýÀº ±âº» ¶Ç´Â °¡Áß ±×·¡ÇÁ¸¸À» ´ë»óÀ¸·Î ¿¬±¸µÇ¾î, ±àÁ¤ ¹× ºÎÁ¤Àû °ü°è¸¦ Æ÷ÇÔÇÏ°í ÀÖ´Â ºÎÈ£Çü ±×·¡ÇÁ¸¦ ´Ù·ç°í ÀÖÁö ¾Ê´Ù.ºÎÈ£Çü ±×·¡ÇÁ´Â ¼Ò¼È ¹Ìµð¾î, ÀüÀÚ »ó°Å·¡»Ó¸¸ ¾Æ´Ï¶ó ±¹¹æ ºÐ¾ßÀÇ ´Ù¾çÇÑ ±º °ü·Ã Á¤º¸µé¿¡¼ ÈçÈ÷ ³ªÅ¸³ª´Â ±àÁ¤ ¹× ºÎÁ¤Àû °ü°èµéÀ» Æ÷ÇÔÇÑ ±×·¡ÇÁÀÌ´Ù. º» ³í¹®¿¡¼´Â ºÎÈ£Çü ±×·¡ÇÁÀÇ °£¼±°ú °£¼±ÀÇ ºÎÈ£±îÁö Àß À¯ÁöÇÏ´Â ºñ¿ëÇÔ¼ö¸¦ »ç¿ëÇÏ¿© Á¤º¸¼Õ½ÇÀ» ÃÖ¼ÒÈÇÏ´Â »õ·Î¿î ±×·¡ÇÁ ¿ä¾à ±â¹ýÀ» Á¦¾ÈÇÏ¿´´Ù. ¶ÇÇÑ, Á¦¾È ¹æ¹ýÀÌ ±âÁ¸ ¹æ¹ý ´ëºñ ÃÖ´ë 24.83% ¿ÀÂ÷°¡ Àû¾ú°í, ºÎÈ£Çü ±×·¡ÇÁÀÇ ±¸Á¶Àû ±ÕÇü ¶ÇÇÑ Àß À¯ÁöÇÔÀ» ½ÇÇèÀûÀ¸·Î º¸¿´´Ù. |
¿µ¹®³»¿ë (English Abstract) |
Recently, as Web services, such as social media and e-commerce, become popular, the size of the graph to express this information is also increasing at an unprecedented rate. In order to efficiently utilize such large graphs, graph summarization methods have been proposed for concise summary graphs with little information loss. However, existing graph summarization methods are limited only to plain graphs or weighted graphs, and they cannot be applied to signed graphs containing both positive and negative relationships. Signed graphs have both positive and negative relations, which are common in social media, e-commerce sites, and various military-related information. In this paper, we propose a new graph summarization technique for signed graphs that produces a signed summary graph using the number of edges and signs incorrectly reconstructed from a summary graph as a cost function. Through experiments, we demonstrate that our method achieves up to 24.83% smaller error while maintaining the structural balance of signed graphs better than competitors. |
Å°¿öµå(Keyword) |
BERT
µö·¯´×
ÅؽºÆ® ºÐ¼®
Deep Learning
Text Analysis
±×·¡ÇÁ ¿ä¾à
ºÎÈ£Çü ±×·¡ÇÁ
±¸Á¶Àû ±ÕÇü
Graph Summarization
Signed Graph
Structural Balance
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|