• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ³í¹®Áö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ³í¹®Áö > Çѱ¹Á¤º¸°úÇÐȸ ³í¹®Áö > Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) Transformer ±â¹Ý ¹ø¿ª¹® »çÈÄ ±³Á¤
¿µ¹®Á¦¸ñ(English Title) Transformer-based Automatic Post-Editing for Machine Translation
ÀúÀÚ(Author) ½ÅÀçÈÆ   ±è¿µ±æ   ÀÌÁ¾Çõ   Jaehun Shin   Youngkil Kim   Jong-hyeok Lee  
¿ø¹®¼ö·Ïó(Citation) VOL 25 NO. 01 PP. 0064 ~ 0069 (2019. 01)
Çѱ۳»¿ë
(Korean Abstract)
¹ø¿ª¹® »çÈÄ ±³Á¤Àº ±â°è ¹ø¿ª ½Ã½ºÅÛÀÇ °á°ú¹°À» ±³Á¤ÇÏ¿© ´õ ³ªÀº ¹ø¿ª¹®À» ¸¸µé¾î³»´Â ¿¬ ±¸·Î, ±â°è ¹ø¿ª ½Ã½ºÅÛ¿¡ µ¶¸³ÀûÀ¸·Î ¹ø¿ª Ç°Áú Çâ»óÀ» À§ÇØ Á¦¾ÈµÈ ºÐ¾ßÀÌ´Ù. º» ³í¹®¿¡¼­´Â ÃÖ±ÙÀÇ ¿¬ ±¸ È帧¿¡ µû¶ó ¹ø¿ª¹® »çÈÄ ±³Á¤À» ¹ø¿ª¹®°ú ¿ø¹®À» º°µµÀÇ ¼Ò½º·Î °£ÁÖÇÏ´Â ´ÙÁß ¼Ò½º ¹ø¿ª ¹®Á¦ (multi-source translation problem)·Î °£ÁÖÇÏ°í, À̸¦ ¹ÙÅÁÀ¸·Î Transformer¸¦ ÀÌ¿ëÇÑ ´ÙÁß ÀÎÄÚ´õ ±¸Á¶ ÀÇ »çÈÄ ±³Á¤ ¸ðµ¨À» Á¦¾ÈÇÑ´Ù. ÀÌ ¸ðµ¨Àº °¢ ÀԷ°ú ±³Á¤¹® »çÀÌÀÇ ÀÇÁ¸¼ºÀ» º°µµ·Î ÇнÀÇÏ°í, À̵éÀ» ´õ ÇÏ¿© ÃÖÁ¾ ÀÔÃâ·Â ÀÇÁ¸¼ºÀ» ¾ò´Â´Ù. WMT17 »çÈÄ ±³Á¤ ¸»¹¶Ä¡¸¦ ÀÌ¿ëÇÑ ½ÇÇè¿¡¼­´Â ´ÙÁß ÀÎÄÚ´õ ¸ðµ¨ÀÌ ±âÁ¸ ¸ðµ¨¿¡ ºñÇØ ´õ ³ªÀº ±³Á¤ ¼º´ÉÀ» º¸¿´À¸¸ç, »çÈÄ ±³Á¤À» °ÅÄ¡Áö ¾ÊÀº ¹®Àå¿¡ ºñÇؼ­µµ ³ªÀº ¹ø¿ª Ç°ÁúÀ» º¸¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Automatic Post-Editing (APE) is a study on the correction for the output of machine translation (MT) systems for the purpose of improving their translation quality. In this paper, APE is regarded as a multi-source translation problem in which one source is an MT output and the other is an input source sentence. So we propose a neural multi-encoder post-editing model based on the transformer. This model aims to find out the dependencies of multiple sources against their post-edited sentence separately. It then adds them to achieve the final input-output dependency. Our experiment on the WMT17 APE data shows that the multi-encoder model outperforms the basic transformer model, producing a better MT output.
Å°¿öµå(Keyword) ±â°è ¹ø¿ª   ¹ø¿ª¹® »çÈÄ ±³Á¤   transformer   ´ÙÁß ÀÎÄÚ´õ ±¸Á¶   machine translation   automatic post-editing   transformer   multi-encoder architecture  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå