æè¿ iPad ã§è«æãç´æ¥èªã¿å§ãã¦ããã®ã ããGoodReader ã§èªãã°ç´ã§èªãã®ã¨ãã»ã©èªå¾æãå¤ãããªããã¨ãåãã£ã¦ããããã¤ã³ãã¯ãã¡ã¢ãä¸ç·ãææ¸ãã§æ¸ãè¾¼ããã¨ãæ¬æã§ãã¼ã¯ãããã¨ããã¯ããã¹ãé¸æãã¦ãã¤ã©ã¤ãããã¨ãããã«ãã¤ã©ã¤ãã§ããã®ã ããããã ã¨è¨æ¶ã«æ®ããªããéã«ãææ¸ãã§æ¸ãè¾¼ãã¨ãå½ç¶ãããã«ç·ã¯å¼ããªãã®ã ãããã¨ã§è¦ç´ãã¦ãããããããã¯ç¢ºãã«èªåã®æ¸ãããã®ã ãã¨åããã®ã§ãé ã«å ¥ããããã®ã§ããã
ææ¸ãã§æ¸ãè¾¼ãã¨ãããã¨ããã®ã¯å æ¥NTTã®ã³ãã¥ãã±ã¼ã·ã§ã³ç§å¦åºç¤ç 究æã«è¡ã£ãã¨ãã«æãã¦ããã£ãã®ã§ããã¯ããããã便å©ãªä½¿ãæ¹ããã¦ãã人ã«èãã¦ã¿ããã®ã§ãããä½ç¹ã¯ãæã§æ¸ãã¨ãªããªãæåãæ¸ããªãã¨ãããªã®ã ãããããiPadç¨ã®ã¹ã¿ã¤ã©ã¹ãè²·ãã°ç¸å½æ¹åããããããã®ã§ãæ©éçºæ³¨ãã¦ã¿ããç´ã§å°å·ãããã®ã«æ¸ãè¾¼ãã§ã¹ãã£ã³ããã¨ãã©ããã¦ãOCR誤ãã¨ã®æ¦ãã«ãªãã®ã ããGoodReaderã§æ¸ãè¾¼ãã°ãå ã®ããã¹ãã¯ãã®ã¾ã¾ã§æ¸ãè¾¼ãã æåãå³ããªã¼ãã¼ã¬ã¤ãããã ããªã®ã§ãæ¤ç´¢æ§ãæãªããªãã®ãããã
ç®ä¸ä¸çªæ©ã¾ããã®ã¯ãMacã®ãã£ã¹ãã¬ã¤ã§è¡¨ç¤ºããPDFã«ææ¸ãã§æ¸ãè¾¼ãã®ãé£ããã¨ãããã¨ãªã®ã ããç®ã®åã«PDFããã£ã¦ãiPadã§èªãã ã»ããèªã¿ãããã£ããããã®ã ãããâ¦â¦ããã£ã±ãç 究室ã«ãããå°å·ãã¦èªãã ã»ããæ軽ãããªãã
ãã¦ãæ¼ããæ©æ¢°ç¿»è¨³ã®åå¼·ä¼ãä»å¹´åº¦ããæ¾æ¬ç 究室ã®æ©æ¢°ç¿»è¨³é¢ä¿ã®ã¡ã¤ã³ã¹ã¿ããã¯Kevinããã«ãªã£ãã®ã§ããããããé¡ãããã次ãã次ã«ã¢ã¤ãã¢ãåºã¦ãã¦ãããããããªããç 究ããããããªãæããGrahamãããéä¸ããé²æã®è©±ããã¦ãã ãã£ãããM1ã®äººãã¡ããã¾ããããã ãã§ããã¨ããã®ã ãã©ãæ¾æ¬ç ã® M1 ã®äººãã¡ã¯ãã§ã«åå¼·ä¼ã ããã§ãå¤åããåå ããã®ã¯é£ããã®ããªâ¦â¦ãæ©æ¢°ç¿»è¨³ãç 究ãã¼ãã«ããããªããä»ããã©ã£ã·ã浸ãã£ã¦ãã£ãã楽ããã¨æãã®ã ãã©ãªãã
Kevinããã¨ãåå¼·ä¼ã®è¨èªã«ã¤ãã¦ã¡ãã£ã¨è©±ããèªåã¨ãã¦ã¯ã1人ã§ãè±èªã®ã»ãããã人ããããè±èªã«ãããã¨ããããªã·ã¼ã«ãã¦ãã (å¥ã®åè£ã¨ãã¦ã¯ã対象ã«ãã¦ããè¨èªãæ¥æ¬èªãªãæ¥æ¬èªãè±èªãªãè±èªããã以å¤ãªãæ¥æ¬èªãè±èªã®ã©ã¡ãã好ããªã»ã) ã®ã ããä»ã®ã¨ããèªåãç´æ¥é¢ãã£ã¦ããåå¼·ä¼ã¯ã©ã¡ããæ¥æ¬èªã§ãã£ã¦ãã¦ããªããªãå ¨é¨è±èªã«ããã®ã¯é£ããã
Kevinããèªèº«ã¯åå¼·ä¼ã§ãæ¥æ¬èªã§ãè±èªã§ããèªåããããããè¨èªã§è©±ãã¦ããããã°ããã大äºãªã®ã¯è¨èªãããªãã¦ãç¸æã¨ã³ãã¥ãã±ã¼ã·ã§ã³ãããã¨ããæ°æã¡ã ãããä¼ããã«ããã£ããåªåãã¦ããããã°ããããä¸å½èªã ã¨ã¡ãã£ã¨å°ããã© (ç¬)ãã¨ããããªã·ã¼ã説æããã¦ãã¦ããããç´ æµã ãªããã¨æã£ãã
ãã ãM1ã®äººãè±èªãåãã£ã¦ãããã©ããæ°ã«ããã¦ãã¦ã"Did I speak too fast?" ãªã©ã¨ããèãããã®ã ããæããM1ã®äººãã¡ããã«ã¼ã³ã¨ãã¦ããããããã¯èªç¶è¨èªå¦çãæ å ±ç³»ã®å°éç¨èªãè±èªã§ç¥ããªãã®ã§æå³ä¸æãªã ãã§ããã£ãã話ãã¦ãå¤ãããªãããããªãããªã¨æããããã¯å¤§å¤æ°ã®äººãå¦é¨ã®ããã¨å°éåéãå¤ãã¦ããNAISTç¹æã®äºæ ãããããªãããæåãåå¹´ãããã¯å¨è¾ºã§è©±ããã¦ããä¼è©±ãæ¥æ¬èªã§ãã£ã¦ãã¡ãã·ãããã·ãã§ããã°ãããã®åéã®åºç¤ç¥èãã¤ãã¦ããã§ãªãã¨ãæå³åãããªãã ããããã®æ¥è¨ã§ãã¨ãã©ãæ¸ãã¦ããããèªåãM1ã®ã¨ãè±èªã§æ¸ããã8ãã¼ã¸ã®è«æãèªãã®ã«50æéããã£ãã®ã§â¦â¦ã
å¤æ¹ã¯5æ28æ¥éå¬ã®ãªã¼ãã³ãã£ã³ãã¹ã®æºåãã¼ãã£ã³ã°ãä»å¹´ããã¢ã¼æ å½ã«ãªã£ããå»å¹´ã¯æããæ©ã¾ã§æ»ã¬ã»ã©å¿ããã£ããããªè¨æ¶ãèã£ã¦ããâ¦â¦ã
å¤ã¯
- Xiaohua Liu, Kuan Li, Ming Zhou, Zhongyang Xiong. Collective Semantic Role Labeling for Tweets with Clustering. IJCAI 2011.
ãç´¹ä»ãTwitter ã®ããã¹ãã«æå³å½¹å² (è¿°èªé æ§é ) ãä»ä¸ãããã¨ããã ãã ã¨åã«ã¸ã£ã³ã«ãå¤ãã£ãã ãã®è©±ãªã®ã ããtweet ã¯ãããããçããæ å ±éãå°ãªãã®ã§ãä¼¼ã¦ããã¤ã¶ãããã¯ã©ã¹ã¿ãªã³ã°ãã¦ãã¾ã¨ãã¦ã¿ã°ä»ããããã¨ããã¨ãããã¡ãã£ã¨æ®éã®è¨å®ã¨éã (ä¸è¬çã«ã¯å ¥åæãã¨ã«ã¿ã°ä»ããã)ãTwitter 解æã£ã½ãã¦ããããªã¨æã£ãã®ã§ããã(ã¾ããã¢ã¤ãã¢ä¸çºã§ããã¨ã¯ãããªã«è¦æã¯ãªãã®ã ãâ¦â¦ãä¸å¿ Markov Logic Network ã使ãææ³ã¨ãæ¯è¼ããããã¦ããããããããè¨ç·´ãã¼ã¿ãéãã®ã§ãé©åãªæ¯è¼ã«ãªã£ã¦ããªã)
ã¾ããã¨ã©ã¼åæã«ããã¨ãææ¡ææ³ã§è§£æã§ããªãã£ããã®ã®å¤§é¨å (60%以ä¸) ãåèªåå²ãããã¯åè©ä»ä¸ã®èª¤ãã«èµ·å ãããã®ã§ãããã¨ãããã¨ã§ãè±èªã§ããããããåèªåå²ãåè©ä»ä¸ãéè¦ã ãããããããæ¥æ¬èªãããã¨ããå½¢ã§ãã½ã¼ã·ã£ã«ã¡ãã£ã¢è§£æã¨ãã£ã¦è¯ã ãããã¼ã¿ãã¤ãã³ã°ããã人ãå¤ãã®ãããããªãããå°éã«å½¢æ ç´ è§£æã¬ã¤ã¤ã¼ã®ç 究ããããè¾æ¸ãæ´åããããããã¨ã大äºã ãã¨ãããã¨ãä¼ãããã£ãã®ã§ãã¡ããã©ããã£ãã