Entropy-based Pruning of Backoff Language Modelsãèªãã ï¼åèªN-gramã¯ã¨ã¦ãããã§ãã¦ãã¦è¨èªã¢ãã«ã¨ãã¦ã®æ§è½ã¯ããªãã®ãã®ãªã®ã ãï¼ãªãã®æåããããã«ä¸è¦æ¨¡ã大è¦æ¨¡ãªãã¼ã¿ã«é©ç¨ãããã¨ããã¨ãµã¤ãºã馬鹿ã§ãããªã£ã¦ãã¾ãï¼ãã®ããã®å¯¾çã¨ãã¦ããããã®ãèªå½ã®ãµã¤ãºãå¶éããæ¹æ³ã¨ï¼N-gramã®é »åº¦ãä¸å®ä»¥ä¸ã®ãã®ãåãæ¨ã¦ãã¨ããæ¹æ³ï¼å¾è ã®å ´åã¯èªå½ãèªåçã«å¶éãããï¼ï¼Google æ¥æ¬èªN-gramãªã©ãé »åº¦20以ä¸ã®ãã®ãé å¸ããã¦ããï¼å¹çãããã¼ã¿ãµã¤ãºãæ¸ããããã«ã¯é »åº¦ã§ã«ãããªãããæ¹å¼ããã使ããã¦ããã¨æãï¼èªå½ã ãã ã¨ããªãå¶éããªãã¨ãµã¤ãºãæ¸ããªãï¼ï¼ãããã«ãããªãããããã¨æ§è½ã¯ããªãè½ã¡è¾¼ãã®ã§ï¼ãã¾ãæ¹æ³ã¯ãªããã®ãã¨èããããã®ããã®è«æã®ææ³ã§ããï¼N-gramã®ãã¼ã¿ã«ã¯é »åº¦ã®é«ã
{{#tags}}- {{label}}
{{/tags}}