Á¤º¸°úÇÐȸ ³í¹®Áö B : ¼ÒÇÁÆ®¿þ¾î ¹× ÀÀ¿ë
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
ÇüÅÂ¼Ò »çÀÌÀÇ À¯»çµµ¸¦ ÀÌ¿ëÇÑ ¿ë·ÊÀÇ Àǹ̺° ÀÚµ¿ Á¤·Ä |
¿µ¹®Á¦¸ñ(English Title) |
Automatic Conceptual Sorting of Concordances using the Similarity Between Morphemes |
ÀúÀÚ(Author) |
¹é´ëÈ£
ÀÌÈ£
ÀÓÇØâ
¹Úµ¿ÀÎ
Dae-Ho Baek
Ho Lee
Hae-Chang Rim
Dong-In Park
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 25 NO. 01 PP. 0183 ~ 0192 (1998. 01) |
Çѱ۳»¿ë (Korean Abstract) |
¿ë·ÊÀÇ Á¤·ÄÀ̶õ ÄÚÆÛ½º¿¡¼ ÃßÃâµÇ´Â ¿ë·Ê¸¦ Àç¹è¿ÇÏ´Â ÀÛ¾÷À» ¸»ÇÑ´Ù. ±âÁ¸ÀÇ ¿ë·Ê Á¤·Ä ¹æ½ÄÀº ƯÁ¤ ÇüżÒÀÇ »çÀüÀû ¼ø¼¿¡ ÀÇÇÑ Á¤·ÄÀ̾ú±â ¶§¹®¿¡, ¿øÇÏ´Â ¾ð¾î Á¤º¸¸¦ ȹµæÇÏ´Â µ¥´Â ¸¹Àº ¾î·Á¿òÀÌ ÀÖ´Ù. º» ³í¹®¿¡¼´Â ÄÚÆÛ½º¿¡¼ ÃßÃâµÇ´Â ¿ë·Ê¸¦ ÇüżÒÀÇ »çÀüÀû ¼ø¼°¡ ¾Æ´Ï¶ó, Á߽ɾîÀÇ Àǹ̿¡ µû¶ó Á¤·ÄÇÏ°íÀÚ ÇÑ´Ù. ¿ë·Ê¸¦ Á߽ɾîÀÇ Àǹ̺°·Î Á¤·ÄÇϱâ À§Çؼ ¿ë·Ê »çÀÌÀÇ ÀÇ¹Ì À¯»çµµ¸¦ °è»êÇÏ°í, À¯»çÇÑ ¿ë·ÊµéÀ» °°Àº Ŭ·¯½ºÅÍ·Î ¸ðÀ¸´Â °èÃþÀû Ŭ·¯½ºÅ͸µ ±â¹ýÀ» »ç¿ëÇÑ´Ù. ±×¸®°í ¿ë·Ê »çÀÌÀÇ ÀÇ¹Ì À¯»çµµ¸¦ °è»êÇϱâ À§Çؼ´Â, °°Àº ÇüżҰ¡ ³ªÅ¸³ª´Â ºóµµ¿Í ÇüÅÂ¼Ò »çÀÌÀÇ À¯»çµµ¸¦ ÀÌ¿ëÇÑ´Ù. ÇüÅÂ¼Ò »çÀÌÀÇ À¯»çµµ ôµµ·Î´Â »óÈ£ Á¤º¸, »óÈ£ Á¤º¸ À¯»çµµ, ±×¸®°í º¤ÅÍ À¯»çµµ¸¦ »ç¿ëÇÑ´Ù. Ç°»ç űëµÈ ¾à 17¸¸ ÄÚÆÛ½º¿¡¼ ÀÇ¹Ì ÁßÀǼºÀÌ ÀÖ´Â ¸í»ç 4°³¿Í µ¿»ç 4°³¸¦ Á߽ɾî·Î »ç¿ëÇÏ¿© ÃßÃâµÈ ¿ë·Ê¿¡ ´ëÇؼ °¢ ¹æ¹ýÀ» ½ÇÇèÇÑ °á°ú, ÇüÅÂ¼Ò »çÀÌÀÇ À¯»çµµ¸¦ »óÈ£ Á¤º¸¿Í »óÈ£ Á¤º¸ À¯»çµµ¸¦ »ç¿ëÇÑ ½ÇÇèÀÌ 90.16%ÀÇ Á¤È®µµ¸¦ º¸¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
A concordance sorting is the procedure of reordering concordances extracted from corpus. The previous methods of concordance sorting have some problem in acquiring linguistic information because they order concordances by lexicographical order of the specific morphemes. In this paper, we propose a method of ordering the concordances extracted from corpus by the meanings of keywords. To order concordances by the meanings of their keywords, we compute the sense similarity between concordances, and use a hierarchical clustering method to collect conceptually similar concordances in the same cluster. We use the frequency of cooccurring morphemes and the similarity between morphemes to compute the similarity between concordances. Also, we use mutual information, the similarity between mutual information values, and vector similarity for the measure of the similarity between morphemes. We have experimented on each method with the concordances of 4 polysemous nouns and 4 polysemous verbs extracted from 170,000 word size part-of-speech tagged corpus. The method of using both mutual information and the similarity between mutual information values shows 90.16% precision.
|
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|