Scribd
Upload
120 views

NLP Endsem 2015

This document contains 17 questions related to various topics in natural language processing and machine translation. The questions cover topics such as word sense disambiguation, Jensen's inequality, Bayesian statistical machine translation, EM algorithm, latent semantic analysis, parsing with context-free grammars, word alignment in machine translation, and probabilistic context-free grammars. The examinee is instructed to answer the questions concisely in the allotted 3 hours.

Uploaded by

Puneet Sangal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
120 views

NLP Endsem 2015

This document contains 17 questions related to various topics in natural language processing and machine translation. The questions cover topics such as word sense disambiguation, Jensen's inequality, Bayesian statistical machine translation, EM algorithm, latent semantic analysis, parsing with context-free grammars, word alignment in machine translation, and probabilistic context-free grammars. The examinee is instructed to answer the questions concisely in the allotted 3 hours.

Uploaded by

Puneet Sangal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

CS 6370 (NLP) – Endsem – Nov 21, 2012

Time: 3 hours

(Be concise. Marks may be deducted for answers that are unnecessarily verbose.)

1. Explain the central idea behind dominance based Word Sense Disambiguation using
a concrete example. [2]

2. What is Jensen’s inequality? Using Jensen’s inequality, show that K-L divergence is
always non-negative. [1+2]

3. Provide a Bayesian setting for Statistical Machine Translation by giving an


interpretation for the likelihood and prior terms. How are these terms estimated in
practice? [2+2]

4. Identify one central limitation of the EM algorithm, and one way to address this
limitation. [1]

5. What is the idea behind transfer-based Machine Translation? Give an example. [2]

6. Context sensitive spelling correction can be viewed as a classification problem.


Explain briefly one technique for performing feature selection in this context. [2]

7. Propose one path based measure of Wordnet based similarity. Show why it works and
the properties it satisfies. How can Information Theoretic approaches be used to
better path based approaches? [3]

8. Explain precisely the most important reason why (a) parameter estimation in PCFGs
uses EM instead of Maximum Likelihood estimation using direct counts from corpus
(b) finite state transducers are used for lemmatization when dictionaries carrying
derivation/inflection information are available (c) empirical NLP has dominated over
rationalistic (classical) NLP over the last two decades (d) interlingua is an attractive
option when machine translators have to be built between several pairs of languages.
[2]

9. List properties of hanger, stretcher and aligner matrix. What is the interpretation of
these matrices in the context of LSI? Explain clearly the geometry of SVD using
these matrices. [5]

10. Define rank of a matrix. In LSI, how does rank reduction correspond to concept
extraction? Explain using two limiting cases, one of a full rank matrix and another of
a maximally rank deficient one. [2]

11. Draw a picture to illustrate a situation where LSA can fail to extract concepts, but
PLSA may succeed. Justify in a single sentence your answer. [1]

12. Discuss very briefly a bootstrapping approach to Word Sense Disambiguation. [2]

13. In the context of PLSA, identify the parameters that need to be estimated, the
objective function and the constraints. Instead of using a conventional optimization
technique like hill climbing, why is EM used? [3]
14. Explain using an example, the following ideas: (a) Lexical Chains (b) Explicit
Semantic Analysis (c) smoothing in the context of Language Models (d) inferencing
step of Information Extraction (e) HMMs for sequence modeling (f) dynamic
programming in parsing. [6]

15. Consider a Machine Translation parallel corpus having three sentence pairs. The
first sentence pair is “come here fast”/”jaldi idhar aao”. The second sentence pair is
“come here”/”idhar aao”. The third sentence pair is “come”/”aao”. (a) Show how the
first few iterations of EM are useful in learning word alignments from this corpus.
Make clear any simplifying assumptions on top of IBM Model 3. (b) How is extra
knowledge “getting generated” in successive iterations of EM? [5+1]

16. What limitations of the basic parsing techniques does the CYK parser address? Is
there an assumption on the grammar rules that CYK can deal with? If yes, what are
these? Given the grammar below and the input sentence “ w =(()(()))”, show the
steps in chart parsing using CYK. Alongside your charts showing each step, mention
clearly the rule(s) that is(are) used (if any) to advance to this step from the previous
one. [4]
S → SS
S →(S1
S1 → S)
S → ()

17. A PCFG is based on the following rules:


a. S → A B
b. B → D A
c. B → D A C
d. A → A C
e. A → a
f. A → b c
g. A → b d e
h. C → f g h
i. D → i
The corpus has the following two sentences, the first occurring 15 times and the
second 30 times:
1. a i b c f g h
2. b c i b d e
(a) Are the sentences accepted by the grammar? In case both of them are, which of
these two sentences is/are ambiguous? Show all possible parse trees of the
sentence(s).
(b) Make an APPROPRIATE initial choice of the rule probabilities. Show the first
three steps of the EM algorithm for estimating the parameters of this PCFG. [5]
END

You might also like