Inf Course Note - Accelerated Natural Language Processing
爱丁堡大学信息学院课程笔记 Accelerated Natural Language Processing, Informatics, University of Edinburgh References: Accelerated natural language processing ANLP revision guide Lecture Slides from the Stanford Coursera course Natural Language Processing, by Dan Jurafsky and Christopher Manning 概率模型 Probability Model 概率模型是随机现象的数学表示,由样本空间,样本空间内的事件以及与每个事件相关的概率定义。目标是模拟给一个事件发生的概率 估算概率(Probability Estimation)一般使用最大似然估计(MLE,相关频率):$$p(x_i) = \frac{Count(x_i)}{\sum_{i=0}^nCount(x_i)}$$ 平滑Smoothing 一般用于处理0概率的问题,比如在训练集中看不到, 但出现在测试集中的词。 Language modeling To compute the probability of sentence /sequence of words $P(w_1, w_2, w_3…)$, or to predict upcomming words $P(w|w_1, w_2, w_3…)$… a language model is also a probability model....