LATE 2 SS11 : SMT EXAM PREPARATION QUESTIONS

1) Please formulate the SMT approach with Bayes Decision Rule.

2) What is the workflow of a SMT system with dependencies of their
components?

3) How to derive the word alignment model: IBM model-1, IBM model-2 and HMM
model? Which of them is/are zero-order model and which of them is/are
first-order model?

4) Which algorithm do you know to estimate parameters for word alignment
model? How to derive the parameters of IBM model-1, IBM model-2 and HMM
model, respectively.

5) What is the main features of IBM model-3, IBM model-4 and IBM model-5?
Explain the reason to introduce those models.

6) How to calculate a LM perplexity? What is an n-gram LM?

7) Which types of LM smoothings are there? Shortly explain linear
discounting and absolute discounting, with linear interpolation and
backing-off, respectively, please write down their equations.

8) Explain the CKY algorithm for parsing.

9) What are top-down parsing and bottom-up parsing?

10) What is forward-backward algorithm, e.g. for HMM?

11) Algorithm and complexities of non-monotone DP search and monotone DP
search.

12) Re-ordering constraints of search.

13) Phrase extraction algorithm

14) ITG search.