Á¤º¸°úÇÐȸ ³í¹®Áö B : ¼ÒÇÁÆ®¿þ¾î ¹× ÀÀ¿ë
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
°È ÇнÀ¹ýÀ» ÀÌ¿ëÇÑ È¿°úÀûÀÎ ÀûÀÀÇü ´ëÈ Àü·« |
¿µ¹®Á¦¸ñ(English Title) |
An Effective Adaptive Dialogue Strategy Using Reinforcement Learning |
ÀúÀÚ(Author) |
±è¿øÀÏ
°í¿µÁß
¼Á¤¿¬
Wonil Kim
Youngjoon Ko
Jungyun Seo
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 35 NO. 01 PP. 0033 ~ 0040 (2008. 01) |
Çѱ۳»¿ë (Korean Abstract) |
Àΰ£Àº ´Ù¸¥ »ç¶÷°ú ´ëÈÇÒ ¶§, ½ÃÇàÂø¿À °úÁ¤À» °ÅÄ¡¸é¼ »ó´ë¹æ¿¡ °üÇÑ ÇнÀÀÌ ÀϾÙ. º» ³í¹®¿¡¼´Â ÀÌ·± °úÁ¤ÀÇ °ÈÇнÀ¹ý(Reinforcement Learning)À» ÀÌ¿ëÇÏ¿© ´ëȽýºÅÛ¿¡ ÀûÀÀÇü ´É·ÂÀÇ ºÎ¿© ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. ÀûÀÀÇü ´ëÈ Àü·«À̶õ ´ëȽýºÅÛÀÌ »ç¿ëÀÚÀÇ ´ëÈ Ã³¸® ½À¼ºÀ» ÇнÀÇÏ¿©, »ç¿ëÀÚ ¸¸Á·µµ¿Í È¿À²¼ºÀ» ³ôÀÌ´Â °ÍÀ» ¸»ÇÑ´Ù. °È ÇнÀ¹ýÀ» È¿À²ÀûÀ¸·Î ´ëÈó¸® ½Ã½ºÅÛ¿¡ Àû¿ëÇϱâ À§ÇÏ¿© ´ëȸ¦ ÁÖ´ëÈ¿Í ºÎ´ëÈ·Î ³ª´©¾î Á¤ÀÇÇÏ°í »ç¿ëÇÏ¿´´Ù. ÁÖ´ëÈ¿¡¼´Â ÀüüÀûÀÎ ¸¸Á·µµ¸¦, ºÎ´ëÈ¿¡¼´Â ¿Ï·á ¿©ºÎ, ¿Ï·á½Ã°£, ¿¡·¯ Ƚ¼ö¸¦ ÀÌ¿ëÇؼ ½Ã½ºÅÛÀÇ È¿À²¼ºÀ» ÃøÁ¤ÇÏ¿´´Ù. ¶ÇÇÑ ÇнÀ °úÁ¤¿¡¼ÀÇ »ç¿ëÀÚ ÆíÀǼºÀ» À§ÇÏ¿© ½Ã½ºÅÛ »ç¿ë ¿ª·®¿¡ µû¶ó »ç¿ëÀÚ¸¦ µÎ ±×·ìÀ¸·Î ºÐ·ùÇÑ ÈÄ ÇØ´ç ±×·ìÀÇ °È ÇнÀ ÈÆ·Ã Á¤Ã¥À» Àû¿ëÇÏ¿´´Ù. ½ÇÇè¿¡¼´Â °³Àκ°, ±×·ìº° °È ÇнÀ¿¡ µû¶ó Á¦¾ÈÇÑ ¹æ¹ýÀÇ ¼º´ÉÀ» Æò°¡ÇÏ¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
In this paper, we propose a method to enhance adaptability in a dialogue system using the reinforcement learning that reduces response errors by trials and error-search similar to a human dialogue process . The adaptive dialogue strategy means that the dialogue system improves users¡¯ satisfaction and dialogue efficiency by learning users¡¯ dialogue styles. To apply the reinforcement learning to the dialogue system, we use a main-dialogue span and sub-dialogue spans as the mathematic application units, and evaluate system usability by using features; success or failure, completion time, and error rate in sub-dialogue and the satisfaction in main-dialogue. In addition, we classify users¡¯ groups into beginners and experts to increase users¡¯ convenience in training steps. Then, we apply reinforcement learning policies according to users¡¯ groups. In the experiments, we evaluated the performance of the proposed method on the individual reinforcement learning policy and group¡¯s reinforcement learning policy.
|
Å°¿öµå(Keyword) |
´ëÈ ½Ã½ºÅÛ
ÀûÀÀÇü ´ëÈ Àü·«
°È ÇнÀ
ÁÖ´ëÈ¿Í ºÎ´ëÈ
Q-ÇнÀ¹ý
Dialogue System
Adaptive Dialogue Strategy
Reinforcement Learning
Main-dialogue and Sub-dialogue
Q-learning
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|