• 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) ÁØ-Áöµµ ÇнÀÀ» ÀÌ¿ëÇÑ »èÁ¦ ±â¹Ý Çѱ¹¾î ¹®Àå Ãà¾à
¿µ¹®Á¦¸ñ(English Title) Deletion Based Korean Sentence Compression with Semi-Supervised Learning
ÀúÀÚ(Author) ¹ÚÁظ𠠠³ëÀ±¼®   ¹Ú¼¼¿µ   Junmo Park   Yunseok Noh   Seyoung Park  
¿ø¹®¼ö·Ïó(Citation) VOL 26 NO. 11 PP. 0513 ~ 0518 (2020. 11)
Çѱ۳»¿ë
(Korean Abstract)
¹®Àå Ãà¾àÀº ÀÚ¿¬¾î ¹®Àå¿¡¼­ ºÒÇÊ¿äÇÑ Á¤º¸¸¦ »èÁ¦ÇÏ°í ÇÊ¿äÇÑ Á¤º¸¸¸À» ³²°Ü ¿øº» ¹®À庸´Ù ª°í °£·«ÇÏ°Ô ¸¸µå´Â °ÍÀÌ´Ù. ÃÖ±Ù ¹®Àå Ãà¾à ºÐ¾ß¿¡¼­ ½ÉÃþ ½Å°æ¸ÁÀ» ÀÌ¿ëÇÑ ¿¬±¸µéÀÌ È°¹ßÇÏ°Ô ¼öÇàµÇ°í ÀÖ´Ù. ÇÏÁö¸¸ ÇнÀÀ» À§ÇÑ Çѱ¹¾î Ãà¾à µ¥ÀÌÅÍ´Â ºÎÁ·ÇÏ´Ù. º» ³í¹®¿¡¼­´Â ½ÉÃþ ½Å°æ¸ÁÀ» ÀÌ¿ëÇÑ ¹®Àå Ãà¾àÀÇ ºÎÁ·ÇÑ µ¥ÀÌÅ͸¦ ±Øº¹Çϱâ À§ÇØ ÁØ-Áöµµ ÇнÀ ¹æ½ÄÀ» äÅÃÇÏ¿© ¹®Àå Ãà¾àÀ» ¼öÇàÇÑ´Ù. ÇнÀÀ» À§ÇØ ´ë·®ÀÇ À¥ »óÀÇ ´º½º·ÎºÎÅÍ µ¥ÀÌÅ͸¦ ÀÚµ¿ ±¸ÃàÇÏ°í, »ç¶÷ÀÌ Á÷Á¢ ±¸ÃàÇÑ °íÇ°ÁúÀÇ µ¥ÀÌÅÍ·Î ¸ðµ¨À» ÆÄÀÎ-Æ©´× ÇÑ´Ù. ÇнÀµÈ ¸ðµ¨Àº ¿øº» ¹®ÀåÀ¸·ÎºÎÅÍ ¿ä¼Ò¸¦ »èÁ¦ °áÁ¤ÇÏ´Â ÀÌÁø ½ÃÄö½º¸¦ Ãâ·ÂÇÏ¿© ¹®ÀåÀ» Ãà¾àÇÑ´Ù. ¸ðµ¨À» ÅëÇØ ¸¸µé¾îÁø Ãà¾à ¹®Àå¿¡ ´ëÇÑ Æò°¡¸¦ À§Çؼ­ Á¤·®Àû, Á¤¼ºÀû Æò°¡¸¦ ¼öÇàÇÑ °á°ú, ÁØ-Áöµµ ¹æ½ÄÀ» ÅëÇØ ÇнÀÇÑ ¸ðµ¨ÀÌ »ç¶÷ÀÌ Á÷Á¢ ±¸ÃàÇÑ ¼Ò·®ÀÇ µ¥ÀÌÅÍ·Î ÇнÀ½ÃŲ ¸ðµ¨º¸´Ù ´õ ÁÁÀº ¼º´ÉÀ» º¸¿©ÁÖ¾ú´Ù.
¿µ¹®³»¿ë
(English Abstract)
Sentence compression is a task to make shorter and simpler sentences by deleting unnecessary information and leaving necessary information. Recently, studies using deep neural networks have been actively conducted in the field of sentence compression. However, there is a lack of Korean compression data for training. In this paper, to overcome the insufficient data of sentence compression, we adapt semi-supervised learning to sentence compression. For training, data is automatically collected from a large amount of news on the Web, and we fine-tune the models using with high-quality data collected by humans. The trained model compresses a sentence by outputting a binary sequence that decides to delete the elements of the original sentence. As a result of the quantitative and qualitative evaluations to evaluate the compressed sentences through the models, the model through the semi-supervise method showed better performance than the model trained with a small amount of data constructed by humans.
Å°¿öµå(Keyword) Çѱ¹¾î ¹®Àå Ãà¾à   ÁØ-Áöµµ ÇнÀ   »èÁ¦ ±â¹Ý Ãà¾à   ÃßÃâ Ãà¾à   µ¥ÀÌÅÍ ÀÚµ¿ ±¸Ãà   Korean sentence compression   semi-supervised learning   deletion-based compression   extractive compression   automatic data set construction  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå