• 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¸°úÇÐȸ³í¹®Áö (Journal of KIISE)

Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)

Current Result Document : 31 / 32

ÇѱÛÁ¦¸ñ(Korean Title) Æ®·£½ºÆ÷¸Ó ¸ðµ¨À» ÀÌ¿ëÇÑ Çѱ¹¾î¿¡¼­ÀÇ ¼ýÀÚ Á¤±ÔÈ­
¿µ¹®Á¦¸ñ(English Title) Number Normalization in Korean Using the Transformer Model
ÀúÀÚ(Author) õÀçÀ±   Á¶Âù¼Û   ÀÌÁ¤ÇÊ   ±¸¸í¿Ï   Jaeyoon Chun   Chansong Jo   Jeongpil Lee   Myoung-Wan Koo  
¿ø¹®¼ö·Ïó(Citation) VOL 48 NO. 05 PP. 0510 ~ 0517 (2021. 05)
Çѱ۳»¿ë
(Korean Abstract)
Çѱ¹¾îÀÇ Á¤±ÔÈ­ ÀÛ¾÷Àº À½¼º ÇÕ¼º ½Ã½ºÅÛÀ» À§ÇÑ ÅؽºÆ® Àüó¸® °úÁ¤¿¡¼­ Áß¿äÇÑ ¿ä¼ÒÀÌ´Ù. ƯÈ÷ Çѱ¹¾î¿¡¼­ ¼ýÀÚ´Â ¹®¸Æ ¿ä¼Ò¿¡ ÀÇÇØ ´Ù¾çÇÏ°Ô ÀÐÈ÷¹Ç·Î ¼ýÀÚ¸¦ Çѱ¹¾î·Î ¹Ù²Ù´Â Á¤±ÔÈ­ ±â¼úÀÇ ¼º´ÉÀÌ ½Ã½ºÅÛÀÇ ¼º´É°ú Á÷°áµÈ´Ù. ±×·¯³ª ÀÌ¿Í °°Àº Çѱ¹¾î¿¡¼­ÀÇ ¼ýÀÚ Á¤±ÔÈ­´Â ³íÀÇµÈ ¹Ù°¡ ¸¹Áö ¾ÊÀ¸¸ç ±âÁ¸ ¸ðµ¨Àº ±ÔÄ¢¿¡ ±â¹ÝÇÏ°í ÀÖ¾î ´Ù¾çÇÑ ¸Æ¶ô¿¡¼­ÀÇ ¼ýÀÚ¸¦ Çѱ¹¾î·Î Á¤±ÔÈ­ Çϴµ¥ ÇѰ踦 º¸ÀδÙ. ÀÌ¿¡ º» ³í¹®Àº µö·¯´×À» ±â¹ÝÀ¸·Î ÇÑ Çѱ¹¾î¿¡¼­ÀÇ ¼ýÀÚ Á¤±ÔÈ­ ¸ðµ¨À» Á¦¾ÈÇÑ´Ù. Á¦¾È ¸ðµ¨Àº ¹®ÀåÀÇ À½ÀýÀ» ÀÔ·ÂÀ¸·Î ÇÏ´Â ½ÃÄö½º Åõ ½ÃÄö½º Æ®·£½ºÆ÷¸Ó ¸ðµ¨À» »ç¿ëÇÏ¿´À¸¸ç ±ä ¼ýÀÚ¿¡ ´ëÇÑ Á¤º¸·Î½á ¼ýÀÚ ÀÚ¸´¼ö ÀÎÄÚµùÀ» °áÇÕÇÏ¿´´Ù. ¼º´É ºñ±³¸¦ À§ÇØ ÀϹÝÀûÀÎ ¼ýÀÚ, »õ·Î¿î ¼ýÀÚ, ºñÇ¥ÁØÀû ¼ýÀÚ, ±ä ¼ýÀÚ Å×½ºÆ® ¼ÂÀ» »ç¿ëÇÏ¿© ½ÇÇèÇÏ¿´´Ù. ±× °á°ú Á¦¾È ¸ðµ¨ÀÌ ±ÔÄ¢ ±â¹Ý ¸ðµ¨°ú ºñ±³ÇØ ÀÏ¹Ý Å×½ºÆ®¼Â¿¡¼­ 2%, ºñÇ¥ÁØÀû Å×½ºÆ®¼Â¿¡¼­ 19% ÀÌ»óÀÇ ¼º´É Çâ»óÀÌ ÀÖ¾ú´Ù. ¶ÇÇÑ ¼ýÀÚ ÀÚ¸´¼ö ÀÎÄÚµùÀ» °áÇÕÇÑ Á¦¾È ¸ðµ¨ÀÌ ´Ù¸¥ µö·¯´× ¸ðµ¨¿¡ ºñÇØ ±ä ¼ýÀÚ Á¤±ÔÈ­¿¡ 13% ³ôÀº ¼º´ÉÀ» º¸¿´´Ù
¿µ¹®³»¿ë
(English Abstract)
Text normalization is a significant component of text-to-speech (TTS) systems. Since numbers in Korean are read in various ways according to their context, number normalization in Korean is crucial to improving the quality of TTS systems. However, the existing model is based on ad hoc rules that are inappropriate for normalizing non-standard numbers. The purpose of this study was to propose a model of number normalization in Korean based on the sequence-to-sequence Transformer model. Moreover, number positional encoding was added to the model to handle long numbers. Overall, the proposed model achieved 98.80% f1 score in the normal test dataset and 90.1% in the non-standard test dataset, which were 2.52% and 19% higher, respectively, than the baseline model. In addition, the proposed model demonstrated a 13% improvement in the longer-number test dataset compared to the other deep learning models.
Å°¿öµå(Keyword) ÅؽºÆ® Á¤±ÔÈ­   ¼ýÀÚ Á¤±ÔÈ­   ½ÃÄö½º Åõ ½ÃÄö½º   Æ®·£½ºÆ÷¸Ó   À¥ ÅؽºÆ® µ¥ÀÌÅÍ   text normalization   number normalization   sequence to sequence   transformer   web text data  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå