MIRAã¨ããã¢ã«ã´ãªãºã ãçµ±è¨çä¿ãåãã®å¦ç¿ã§ããæ績ãå©ãåºãã¦ããããã§ãï¼ ä¿ãåãã«ç¹åããã¢ã«ã´ãªãºã ã§ã¯ãªãï¼structured output ãªãã»ã¼ä½ã§ãã§ããé常ã«æ±ç¨æ§ã®é«ãã¢ã«ã´ãªãºã ã®ããã§ãï¼è©³ç´°ã¯ãã¡ã é¢ç½ãããªã®ã§ï¼ã¡ãã£ã¨æ·±è¿½ããã¦ã¿ã¾ããï¼ç¹å¾´ãã¾ã¨ãã㨠- ãªã³ã©ã¤ã³å¦ç¿ - k-best解ãå¾ããããã㪠decoder ããããã°åã - single-best ã§ããã¡ããå¯è½ - single best ã®å ´å㯠Collins voted perceptron ã«é ·ä¼¼ - single best ã®å ´åã® inference 㯠SMO ã¨å ±éç¹ãããï¼å®é max-margin parsing ã®ç¹æ®ç³»ã«ãªã£ã¦ãã ãªã©ãªã©ï¼é¢ç½ãç¹ãããããããã¾ãï¼ ãã¨ãã¨ã¯ Ben Tasker ã® Max margin parsing ã®
{{#tags}}- {{label}}
{{/tags}}