課程名稱︰自然語言處理 NATURAL LANGUAGE PROCESSING
課程性質︰資訊所選修／知識管理學程選修
課程教師︰陳信希
開課系所︰資訊所
考試時間︰2005 Jan 13
試題 :
1. What is Bayes' Theorem? What is noisy channel model? Take speech
recognition as an example to describe the concepts. (15 points)
2. What is mutual information? Please describe its pysical meanings.
(15 points)
3. What is collocation? Please show the t test procedure to acquire
collocations. (15 points)
4. What is data sparseness problem? Propose a method for discounting
and model combination approaches, respectively. (15 points)
5. Prepositional phrase attachment aims to determine which constituent
a preposition phrase modifies. Please show a statistical method to
deal with this problem. (10 points)
6. For training Hidden Markov Model, forward probability and backward
probability are defined. Please show the definitions of these two
probabilities and discuss how they are employed to train HMM.
(15 points)
7. Based on your discussion in question 6, please show how to extend
forward and backward probabilities to outside and inside
probabilities. Similarly, please discuss how to train probabilistic
context free grammars. (bonus, 15 points)
8. In our term project, you are asked to design a word sense
disambiguation system. Please describe the approach you adopted
in this project. (15 points)
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 140.112.241.81