Previous Entry Share Next Entry
My easiest exam out of the lot of them.....
Natural Language Processing
Candidates should answer ONLY TWO questions. The time allowed is ONE AND A HALF hours. Each question is worth 20 marks.

1. (a) You are given the following context-free grammar (CFG):

S --> NP VP
NP --> Art N
VP --> V NP
VP --> VP PP
VP --> V NP PP
PP --> P NP
NP --> John
NP --> Mary
Art --> a
N --> book
P --> to
V --> gave

Explain, with diagrams, how the sentence John gave a book to Mary would be parsed by a chart parser using this grammar [10 Marks]

(b) Draw the set of parse trees assigned to this sentence by the grammar [3 Marks]

(c) Describe an extension to CFG that allows parses to be disambiguated. How would it work for the problem at hand? [7 Marks]

2. TotallyNLP Limited is designing a Customer Relationship Management software package, one of whose modules will route email to different destinations according to the language the sender is writing in. The package will be sold to multinational companies who need to deal with customer enquiries in several different languages. A special feature of the module will be that it makes its routing decisions only on the text in the body of the message, not on any information in the header (e.g. country code of the sender's e-mail address) since this may be unreliable.

(a) Describe two distinct techniques that the module could use, paying particular attention to the types of language data that would be required [12 Marks]

(b) Discuss the strengths and weaknesses of each technique. [8 Marks]

3. (a) Briefly describe and give examples of (1) reference ambiguity, and (2) word sense ambiguity. [8 Marks]

(b) In a language analysis module, why would an application developer wish to avoid having to resolve these types of semantic ambiguity? [4 Marks]

(c) For each of two types of natural language application of your choice discuss the extent to which reference ambiguity and word sense ambiguity need to be resolved. [8 Marks]

  • 1
1a. no
1b. no
1c. no
2a. no
2b. no
3a. no
3b. no
3c. no

.... didn't take hour and a half!

Didn't get you many marks either ;o)

hey. (surprise, i'm not commenting re how you look...) i'm into natural language stuff too... all of it more from a cognitivist standpoint, though my neural nets don't exist just yet...(am currently working through thesis...) i could probably do what you did except i haven't studied computational linguistics in a long time. could have done it three years ago, yeah. what you do is pretty cool, meaning blx as well as math/ai/nlp/ ;o)

  • 1

Log in

No account? Create an account