见见大魔王

3 Language Modelling3.1 Direct Application of LM: Noisy Channel Model3.2 How to estimate the probabilities3.2.1 General idea: Probability Factorization3.2.2 Try an easy one first: The Unigram Model3.2.3 General idea: N-gram Language Model3.3 Ev

2 Distributed Word Representation2.1 Representation of Text2.2 Global Generalization of Distributed Representation2.3 Distributed Representation: How?2.3.1 Word Embedding2.3.2 How to make neighbors represent words?2.3.3 Solution 1：Matrix Factor

目录1.1 Inverted Index and Boolean retrieval modelLinearly scanningAvoid linearly scanning： index the documents in advanceConstruct Inverted Index1.2 Rank RetrievalVectorization with TF-IDF weightingBetter scoring functions: BM251.3 PageRank

0.1 什么是NLP？NLP = NLU + NLGNLU：语音/文本 ——>语义 （text——>structure）NLG：语义——>文本/语音 （text——>structure）The challenge : Mutilple ways to express.The challenge : Ambiguity（一词多义）​ 解决方法：尽可能使用上下文信息今天参观了苹果公司现在正好是苹果收获的季节The challenge : Multi-mod