VASAVI COLLEGE OF ENGINEERING (AUTONOMOUS), HYDERABAD – 500 031

DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

B.E VIII - Semester - (2024-25) ECE Branch - I - Internal Test
Title of the course: Language Models and Course Code: U21PE87XEC
Applications
Maximum Marks:
Date: 01-03-2025 Time: 02.30 pm to 4.00 pm
30

BTL Mapped
Questio Mark
Description of the question (1/2/3/4/
n No. s
5/6) CO PO PSO
Part - A (6 × 1 Marks: 06) Answer all questions
1. What is regular expression? write it’s importance.
1 1 1 1 3
2. If a sentence has 7 words, how many bigrams can be formed? 1 1 1 1 3

3. Calculate the cosine similarity between two vectors A = [1, 2] and B = [2, 4]. 1 1 3 1 3

4. What is the main advantage of RNNs over Feedforward Networks in NLP?

1 1 4 1 3
5. What are the applications of RNN Model? 1 1 4 1 3

6. What is the key advantage of Transformers over RNNs? 1 1 5 1 3

Part - B (04 × 3 Marks: 12) Answer all questions

7. Write and explain each character in Regular Expression for extracting mobile
4 3 1 2 3
numbers from the corpus which starts with 7 or 8 or 9.
8. Given the following corpus, compute the TF-IDF table for the terms present in
the documents. Show step-by-step calculations for Term Frequency (TF),
Inverse Document Frequency (IDF), and TF-IDF values.
Corpus:
4 3 3 2 3
D1:"people watch campusX."
D2: "campusX watch campusX."
D3: "people write comment."
D4: "campusX write comment."
9. Explain the training process of two layer fedforward neural networks.
4 2 4 1 3

Part - C (02 × 06 Marks: 12) Answer all questions

10. Compute the Minimum Edit Distance between the words "execution" and
"insertion" using the Levenshtein Distance algorithm. Clearly list the step-by- 6 3 2 2 3
step transformations required to convert "execution" into "insertion".
11. Explain in detail the problem of vanishing gradients in RNNs. How do LSTMs
6 2 4 3 3
address this issue?
Summary of the percentage for each of the criteria BTL (Blooms Taxonomy Level) from the questions framed.
1. Fundamental Knowledge for level – 1 (recall)): 20%
2. Level – 2 (understand):33.33%
3. Knowledge on application and analysis for level-3 (apply) & 4 (Analyze): 46.67%

Date: Name & Signature of the subject

expert

VASAVI COLLEGE OF ENGINEERING (AUTONOMOUS), HYDERABAD – 500 031

 DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING
B.E VIII - Semester - (2024-25) ECE Branch - I - Internal Test
Title of the course: Language Models and Course Code: U21PE87XEC
Applications
Maximum Marks:
Date: 01-03-2025 Time: 02.30 pm to 4.00 pm
30

BTL Mapped
Questio Mark
Description of the question (1/2/3/4/
n No. s
5/6) CO PO PSO
Part - A (6 × 1 Marks: 06) Answer all questions
1. What is Tokenization in NLP? How many words are in the following sentence
1 1 1 1 3
after tokenization? "Natural Language Processing is amazing!"
2. Given the following unigram probabilities, compute the probability of "the 1 1 1 1 3
cat" using a bigram model:P(the) = 0.1, P(cat | the) = 0.5
3. What does POS tagging do in NLP? 1 1 2 1 3

4. What is the role of embeddings in neural NLP models?

1 1 3 1 3
5. What problem does the vanishing gradient issue cause in RNNs? 1 1 4 1 3

6. What is Self-Attention in Transformers? 1 1 5 1 3

Part - B (04 × 3 Marks: 12) Answer all questions

7. Write and explain each character in Regular Expression for extracting pattern
to match phone-like numbers (e.g., 34-3456, 56-4532) and and for patterns
4 3 1 2 3
like dd-mm-yyyy in the corpus= “India 34-3456 12-05-2007, great 56-4532
11-11-2011, Country 67-8945 12-01-2009”
8. Given the following corpus, compute the TF-IDF table for the terms present in
the documents. Show step-by-step calculations for Term Frequency (TF),
Inverse Document Frequency (IDF), and TF-IDF values.
Corpus:
4 3 3 2 3
D1:"people watch campusX."
D2: "campusX watch campusX."
D3: "people write comment."
D4: "campusX write comment."
9.
What is the loss function used in feedforward neural network for NLP Tasks? 4 2 4 1 3
And explain the How Back propagation help in training deep neural networks?
Part - C (02 × 06 Marks: 12) Answer all questions
10. Compute the Minimum Edit Distance between the words "america " and
"costarica " using the Levenshtein Distance algorithm. Clearly list the step-by- 6 3 2 2 3
step transformations required to convert "america" into "costarica".
11. Explain in detail the importance of Attention mechanism in Recurrent neural
6 2 5 3 3
networks with the required skecthes and equations.
Summary of the percentage for each of the criteria BTL (Blooms Taxonomy Level) from the questions framed.
1. Fundamental Knowledge for level – 1 (recall)): 20%
2. Level – 2 (understand): 33.33%
3. Knowledge on application and analysis for level-3 (apply) & 4 (Analyze): 46.67%

Date: Name & Signature of the subject

expert