隐马尔科夫模型和词性标注讲义.ppt

n -1 6.98 v 00 18.22 n 10 19.87 00 21.65 20 25.89 v 10 20.86 00 23.9 20 27.61 a 00 32.42 01 34.2 10 36.16 d 00 29.38 01 31.16 10 31.38 n 10 44.59 00 44.88 11 46.37 v 10 37.47 11 39.25 12 39.47 p 00 14.62 c 00 12.28 n -1 6.98 v 00 18.22 n 10 19.87 00 21.65 20 25.89 v 10 20.86 00 23.9 20 27.61 a 00 32.42 01 34.2 10 36.16 d 00 29.38 01 31.16 10 31.38 n 10 44.59 00 44.88 11 46.37 v 10 37.47 11 39.25 12 39.47 p 00 14.62 c 00 12.28 n -1 6.98 v 00 18.22 n 10 19.87 00 21.65 20 25.89 v 10 20.86 00 23.9 20 27.61 a 00 32.42 01 34.2 10 36.16 d 00 29.38 01 31.16 10 31.38 n 10 44.59 00 44.88 11 46.37 v 10 37.47 11 39.25 12 39.47 p 00 14.62 c 00 12.28 n -1 6.98 v 00 18.22 n 10 19.87 00 21.65 20 25.89 v 10 20.86 00 23.9 20 27.61 a 00 32.42 01 34.2 10 36.16 d 00 29.38 01 31.16 10 31.38 n 10 44.59 00 44.88 11 46.37 v 10 37.47 11 39.25 12 39.47 p 00 14.62 c 00 12.28 N-Best Search结果 1)收入/n 和/c 生活/n 进一步/d 提高/v 37.47 2)收入/n 和/p 生活/n 进一步/d 提高/v 39.25 3)收入/n 和/c 生活/v 进一步/d 提高/v 39.47 谢谢！ * * 词性标注 P(T|W)=P(W|T)P(T)/P(W) argmaxTp(T|W)=argmaxTp(W|T)p(T) P(W|T)=∏i=1…dp(wi|w1,…,wi-1,t1,…,td) p(wi|w1,…,wi-1,t1,…,td) ≌p(wi|ti) P(T)=∏i=1…dp(ti|t1,…,ti-1) p(ti|t1,…,ti-1)=p(ti|ti-n+1,…,ti-1) 有指导的学习 训练时事先对语料库进行了人工的词性标注，因此在训练时看到了状态（词性），属于VMM，在测试时，只能看到观察值（词序列），因此属于HMM。 应用最大似然估计 p(wi|ti)=cwt(ti,wi)/ct(ti) p(ti|ti-n+1,…,ti-1) =ctn(ti-n+1,…,ti-1,ti)/ct(n-1)(ti-n+1,…,ti-1) 平滑 p(wi|ti)：加1平滑 p(ti|ti-n+1,…,ti-1)：线性差值 用带标记的语料进行训练 Pierre/NNP Vinken/NNP , , 61/CD years/NNS old/JJ ,/, will/MD join/VB the/DT board/NN as/IN a/DT nonexecutive/JJ director/NN Nov./NNP 29/CD ./. Mr./NNP Vinken/NNP is/VBZ chairman/NN of/IN Elsevier/NNP N.V./NNP ,/, the/DT Dutch/NNP publishing/VBG group/NN . . Rudolph/NNP Agnew/NNP ,/, 55/CD years/NNS old/JJ and/CC former/JJ chairman/NN of/IN Consolidated/NNP Gold/NNP Fields/NNP PLC/NNP ,/, was/VBD named/VBN a/DT nonexecutive/JJ director/NN of/IN this/DT British/JJ industrial/JJ conglomerate/NN ./. c(JJ)=7 c(JJ, NN)=4, P(NN|JJ)=4/7 无指导的学习 语料库只是词的序列，没有人工标注词性，是Plain Text。 完全无指导的学习是