The IIT Madras BS Introduction to Natural Language Processing (Intro to NLP) Quiz 1 paper sat on 23 Feb 2025, in the January 2025 term: 30 questions for 50 marks in 120 minutes. Every question is below with its answer. Take it as a timed mock test to be marked, or read it through first.
Consider the following conversation:
User: “It’s quite cold in here.”
Bot: “Yes, the temperature is 20^(o)C.”
User: “Could you help?”
Bot: “What would you like help with?”
Which level of language processing is the chatbot failing to implement effectively?
Lexical
Discourse
Pragmatic
Syntactic
Correct answer
Pragmatic
Statement: “Lexical ambiguity occurs when an individual word has multiple dictionary meanings.”
Yes
No
Correct answer
Yes
What makes the example “Time flies like an arrow” particularly challenging for POS tagging?
The presence of words with ambiguous lexical categories that can lead to multiple syntactic structures.
Difficulty in identifying compound nouns within the sentence.
The sentence contains idiomatic expressions that require semantic understanding.
The use of archaic word forms.
Correct answer
The presence of words with ambiguous lexical categories that can lead to multiple syntactic structures.
Anaphora resolution can be fully solved using only syntactic parsing techniques.
Yes
No
Correct answer
No
Which of the following statements correctly describes the relationship between Finite State Automata (FSA) and Finite State Transducers (FST)?
FSAs are designed solely for recognizing patterns in input strings, while FSTs can recognize patterns and produce corresponding outputs.
Both FSAs and FSTs utilize a single tape for processing input, but FSAs have a fixed number of states while FSTs can dynamically change the number of states during execution.
FSTs can be used to perform computations involving multiple input tapes, whereas FSAs are limited to a single input tape.
FSA can produce outputs based on their input states, while FSTs are only capable of recognizing inputs without generating outputs.
Correct answer
FSAs are designed solely for recognizing patterns in input strings, while FSTs can recognize patterns and produce corresponding outputs.
Which of the following sequences represents a valid order of calculations in the Forward algorithm of HMM?
Calculate state probabilities → Update observation probabilities → Compute final sequence probability
Initialize state probabilities → Calculate joint probabilities → Sum over all possible paths
Calculate emission probabilities → Sum transition probabilities → Normalize final states
Initialize α1(i) → Calculate αt(i) recursively → Sum over final states
Correct answer
Initialize α1(i) → Calculate αt(i) recursively → Sum over final states
Which feature combination would be most effectively handled by CRF but not by HMM for POS tagging?
Current word and its immediate next word
Word length and first character capitalization
Previous tag and current word ending
Current word’s morphological features combined with a window of ±2 words and their POS tags
Correct answer
Current word’s morphological features combined with a window of ±2 words and their POS tags
1-B, 2-C, 3-A, 4-D
1-A, 2-B, 3-C, 4-D
1-B, 2-A, 3-D, 4-C
1-D, 2-C, 3-B, 4-A
Correct answer
1-B, 2-C, 3-A, 4-D
1.5
1.6
1.7
1.8
Correct answer
1.6
In evaluating a dependency parser, the following results were obtained on a test set of 500 words: • 450 words have the correct head
• 425 words have both the correct head and correct label
• 475 words have the correct label
What are the Unlabeled Attachment Score (UAS) and Label Accuracy Score (LS)?
UAS: 90%, LS: 95%
UAS: 90%, LS: 85%
UAS: 85%, LS: 95%
UAS: 85%, LS: 90%
Correct answer
UAS: 90%, LS: 85%
The phonological level of NLP processing is crucial for improving the performance of text-based sentiment analysis systems.
Yes
No
Correct answer
No
In the sentence “I saw her duck,” morphological analysis alone is sufficient to determine whether “duck” is being used as a noun or a verb.
Yes
No
Correct answer
No
Which would be the most appropriate approach for processing millions of search queries in real- time?
Lemmatization
Stemming
Full morphological analysis
Manual word reduction
Correct answer
Stemming
Which approach would be most appropriate for building a dictionary lookup system?
Stemming
Lemmatization
Simple suffix removal
No word reduction
Correct answer
Lemmatization
Rule-based POS tagging always results in accurate tagging because it relies on linguistic rules rather than probabilities.
Yes
No
Correct answer
No
Which of the following demonstrates an “overlapping entity” scenario in NER?
Albert Einstein and Theory of Relativity
Amazon and Amazon River
Tesla and Model S
Shakespeare and Hamlet
Correct answer
Amazon and Amazon River
Consider the following phrase structure rule: NP → (Det) (Adj)* N (PP)*
Which of the following statements about this rule is FALSE?
It allows for multiple consecutive adjectives before the noun.
It requires exactly one determiner before any adjectives.
It permits zero or more prepositional phrases after the noun.
The noun is the only mandatory element in this noun phrase.
Correct answer
It requires exactly one determiner before any adjectives.
Given the sentence “The chef cooked a meal for the guests using fresh ingredients,” identify the Semantic Roles of the sentence.
Agent: Chef, Theme: Meal, Beneficiary: Guests, Instrument: Ingredients
Agent: Meal, Theme: Guests, Beneficiary: Chef, Instrument: Cooking
Agent: Chef, Theme: Ingredients, Beneficiary: Meal, Instrument: Guests
Agent: Ingredients, Theme: Chef, Beneficiary: Meal, Instrument: Cooking
Correct answer
Agent: Chef, Theme: Meal, Beneficiary: Guests, Instrument: Ingredients
Context-Free Grammar (CFG) rules can only generate sentences with a fixed word order.
Yes
No
Correct answer
No
The CKY parsing algorithm requires the grammar to be in Chomsky Normal Form (CNF).
Yes
No
Correct answer
Yes
In the sentence “He gave her a book,” the dependency relationship between “gave” and “book” is labeled as nsubj.
Yes
No
Correct answer
No
Which evaluation metric would you use to measure the proportion of correct dependency relations retrieved by a parser from the total number of correct relations?
Precision
Recall
F1 Score
Accuracy
Correct answer
Recall
Given tables are emission and transition probability tables for a POS tagger.
Emission probability:
| Words | Noun | Verb | Adj |
|---|---|---|---|
| book | 0.4 | 0.3 | 0.1 |
| red | 0.1 | 0.0 | 0.5 |
| jumps | 0.0 | 0.6 | 0.0 |
Transition probabilities:
| POS | Noun | Verb | Adj |
|---|---|---|---|
| Noun | 0.2 | 0.5 | 0.3 |
| Verb | 0.4 | 0.1 | 0.5 |
| Adj | 0.6 | 0.3 | 0.1 |
What is the probability of the sequence: Adj Noun Verb for the words “red book jumps”?
0.0090
0.036
0.0045
0.0036
Correct answer
0.036
Which of the following are applications of clustering in NLP tasks? (Select all that apply)
Grouping similar entities in NER
Unsupervised POS tagging
Sentiment analysis
Word clustering into syntactic categories
Machine translation
Correct answers
Grouping similar entities in NER
Unsupervised POS tagging
Word clustering into syntactic categories
Which characteristics distinguish Morphological Analysis (MA)from POS tagging? (Select all that apply)
MA is independent of syntactic rules, while POS tagging often depends on them
MA may produce multiple analyses for a word while POS tagging produces one
MA requires fewer categories compared to POS tagging
MA assigns functional roles in a sentence, while POS tagging deals with form and structure.
Correct answers
MA is independent of syntactic rules, while POS tagging often depends on them
MA may produce multiple analyses for a word while POS tagging produces one
MA requires fewer categories compared to POS tagging
What are potential issues when implementing stemming? (Select all that apply)
Over-stemming
Under-stemming
Requiring a large dictionary
Creating non-words
Correct answers
Over-stemming
Under-stemming
Creating non-words
Which of the following are true about the relationship between POS tagging, NER, and classification models? (Select all that apply)
Both POS tagging and NER can be framed as classification problems
Naive Bayes and logistic regression can be applied to both tasks
Clustering can be used as a preprocessing step for both tasks
POS tagging always precedes NER in NLP pipelines
Both tasks benefit from considering contextual information
Correct answers
Both POS tagging and NER can be framed as classification problems
Naive Bayes and logistic regression can be applied to both tasks
Clustering can be used as a preprocessing step for both tasks
Both tasks benefit from considering contextual information
Which of the following NLP tasks would benefit from processing at the discourse level? (Select all that apply)
Generating a coherent multi-paragraph story
Summarizing a long document
Resolving pronouns in a conversation
Determining the sentiment of a single word
Identifying the topic shifts
Correct answers
Generating a coherent multi-paragraph story
Summarizing a long document
Resolving pronouns in a conversation
Identifying the topic shifts
Which of the following statements about stemming and lemmatization are true?
Stemming can result in non-linguistic root forms, while lemmatization always produces valid words.
Lemmatization requires a vocabulary or lexicon, whereas stemming does not.
Stemming is computationally more expensive than lemmatization because it involves morphological analysis.
Both stemming and lemmatization aim to reduce words to their root forms but differ in their approaches and accuracy.
Correct answers
Stemming can result in non-linguistic root forms, while lemmatization always produces valid words.
Lemmatization requires a vocabulary or lexicon, whereas stemming does not.
Both stemming and lemmatization aim to reduce words to their root forms but differ in their approaches and accuracy.
In a semantic role labeling evaluation, a model has identified 100 correct semantic roles out of 120 total semantic roles it predicted. What is the precision of the model as a decimal (rounded to two decimal places)?
Correct answer: 0.83
