• 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) Bidirectional LSTM-CRF ±â¹Ý Çѱ¹¾î °³Ã¼¸í ÀνÄÀ» À§ÇÑ À¯»ç Á¢»ç ÀÚÁúÀ» ÀÌ¿ëÇÑ ´Ü¾î Ç¥»ó È®Àå
¿µ¹®Á¦¸ñ(English Title) Extending Word Representations with Predicted Affix Features for Bidirectional LSTM-CRF-based Korean Named Entity Recognition
ÀúÀÚ(Author) Á¤¿¹¿ø   ÀÌÁ¾Çõ   Yewon Jeong   Jong-Hyeok Lee  
¿ø¹®¼ö·Ïó(Citation) VOL 26 NO. 09 PP. 0408 ~ 0413 (2020. 09)
Çѱ۳»¿ë
(Korean Abstract)
º» ³í¹®¿¡¼­´Â ½ÉÃþ ½Å°æ¸Á ±â¹ÝÀÇ Çѱ¹¾î °³Ã¼¸í ÀνÄÀ» À§ÇØ µ¥ÀÌÅͷκÎÅÍ Á¢»ç ÀÚÁúÀ» Ãß·ÐÇÏ¿© ´Ü¾î¿Í À½Àý ´ÜÀ§ Á¤º¸¿¡ °áÇÕÇÏ´Â ½Ç¿ëÀûÀÎ ¸ðµ¨À» Á¦¾ÈÇÑ´Ù. ÃÖ±Ù °³Ã¼¸í ÀÎ½Ä ¿¬±¸¿¡¼­ °¡Àå ¿ì¼öÇÑ ±â¹ýÀ» ÀÌ¿ëÇÑ µÎ °³ÀÇ ¸ðµ¨, Bidirectional Long Short Term Memory-Conditional Random Field (bidirectional LSTM-CRF)¿Í Bidirectional Long Short Term Memory-Convolutional Neural Networks-Conditional Random Field (bidirectional LSTM-CNNs-CRF)¸¦ »ç¿ëÇÏ°í, ±âÁ¸ÀÇ ÀÔ·Â ´Ü¾î Ç¥»ó¿¡ Ç°»ç ÀÓº£µù°ú À¯»ç Á¢»ç ÀÓº£µùÀ» Ãß°¡ÇÏ¿© È®ÀåÇÏ¿´´Ù. Çѱ¹¾î Á¢»ç¿¡´Â Á¢µÎ»ç¿Í Á¢¹Ì»ç°¡ ÀÖ°í ´ëºÎºÐ 1À½ÀýÀΠƯ¼ºÀ» °í·ÁÇÏ¿© µîÀå ºóµµ¼ö¿¡ µû¸¥ °£´ÜÇÑ ÇÊÅ͸µ ¹æ¹ýÀ¸·Î À¯»ç Á¢»ç¸¦ ¿¹Ãø, °³Ã¼¸í ÀÎ½Ä ÀÚÁú·Î È°¿ëÇÏ¿´´Ù. ÀÌ ¸ðµ¨Àº 2016³â ±¹¾îÁ¤º¸Ã³¸®´ëȸ¿Í Çѱ¹ÀüÀÚÅë½Å¿¬±¸¿ø(ETRI)¿¡¼­ ¹èÆ÷ÇÑ °³Ã¼¸í ÀÎ½Ä ¸»¹¶Ä¡¿¡ ´ëÇØ º°µµÀÇ ¿ÜºÎ ÀÚ¿ø ¾øÀÌ ±âÁ¸ ¸ðµ¨º¸´Ù F1 Á¡¼ö°¡ ÃÖ´ë 2.44% Çâ»óµÇ´Â °ÍÀ» º¸¿´´Ù
¿µ¹®³»¿ë
(English Abstract)
We propose a Korean named entity recognition model (NER) which extracts affix features to augment word representations. We build upon two recently proposed models using state-of-the-art NER techniques, namely bidirectional LSTM-CRF and bidirectional LSTM-CNNsCRF, by extending the word embedding with part-of-speech and approximated affix information. Because Korean affixes mainly comprise single-character prefixes or suffixes, we chose to use an inexpensive character level frequency filter to infer the affix information. Our experimental results on the HCLT 2016 and ETRI NER datasets showed up to a 2.44% increase in the F1 score compared to the original models without any additional dictionary or morphological tools
Å°¿öµå(Keyword) °³Ã¼¸í ÀνĠ  Çѱ¹¾î °³Ã¼¸í ÀνĠ  bidirectional LSTM-CRF   ´Ü¾î Ç¥»ó   Named Entity Recognition (NER)   Korean Named Entity Recognition   word representations   bidirectional LSTM-CRF  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå