The IIT Madras BS Introduction to Natural Language Processing (Intro to NLP) Quiz 1 paper sat on 27 Oct 2024, in the September 2024 term: 26 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.
In a whimsical language where adjectives change based on the weather, an FST is designed to transform “happy” into “happyrain” on rainy days and “happysun” on sunny days. If the FST has 5 states for “happy” and 3 states each for weather conditions, what’s the minimum total number of states required?
8
11
13
15
Correct answer
11
An FSA recognizes words that begin with "un" and end with "ed". The state transitions are defined as follows:
States:
Transitions:
For which of the following inputs will the FSA reach the accepting state?
“unlearned”
“undefined”
“unfinished”
All of these
Correct answer
All of these
An FSA recognizes binary strings that end with '01'. The state transitions are defined as follows:
States:
Transitions:
What is the output of this FSA for the input string "1101"?
The string is rejected.
The string is accepted.
The FSA transitions to an error state.
The output is “11”.
Correct answer
The string is accepted.
In a graph-based dependency parser using edge-factored scoring, the following edge scores have been calculated for a sentence:
What is the score of the maximum spanning tree (assuming all other possible edges have lower scores)?
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: 85%
UAS: 90%, LS: 95%
UAS: 85%, LS: 95%
UAS: 85%, LS: 90%
Correct answer
UAS: 90%, LS: 95%
A legal document analysis tool is being developed to identify key clauses. If the tool correctly identifies 80 relevant clauses out of 100 actual relevant clauses, and incorrectly flags 20 irrelevant clauses as relevant, what is the F1 score of the tool?
0.80
0.82
0.85
0.89
Correct answer
0.82
In the sentence “The old Shakespearean Globe Theatre on the banks of the Thames hosted a modern rendition of ’A Midsummer Night’s Dream’”, which phrase is NOT a Named Entity?
Shakespearean
Globe Theatre
Thames
A Midsummer Night’s Dream
Correct answer
Shakespearean
In processing the sentence “After John beat him in chess, Bill lost his Queen,” which NLP level is most crucial for understanding “Queen” ?
Lexical
Semantic
Discourse
Pragmatic
Correct answer
Semantic
Anaphora resolution, which falls under the discourse level of NLP, can be fully solved using only syntactic parsing techniques.
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
What is a primary feature of agglutinative languages?
Morphemes are fused together, making them inseparable.
Morphemes are added like beads on a string, each representing one grammatical feature.
Each word consists of a single morpheme.
Morphology does not exist in these languages.
Correct answer
Morphemes are added like beads on a string, each representing one grammatical feature.
In morphological analysis, what is the primary difference between Item and Arrangement (IA) and Item and Process (IP) approaches?
IA focuses on concatenation, while IP uses rules and allomorphs.
IA is used for synthetic languages, while IP is for analytic languages.
IA is word-based, while IP is morpheme-based.
IA uses paradigms, while IP uses linear sequencing.
Correct answer
IA focuses on concatenation, while IP uses rules and allomorphs.
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
Treebanks are automatically generated without human intervention.
Yes
No
Correct answer
No
In the sentence “He gave her a book,” the dependency relationship between “gave” and “book” is labeled as nsubj.
Yes
No
Correct answer
No
In the sentence “The cat sat on the mat,” the prepositional phrase “on the mat” serves as a dependent of “sat” in a case relation.
Yes
No
Correct answer
No
Which of the following NLP tasks would benefit from processing at the discourse level?
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 are characteristics of Naive Bayes classifiers? (Select all that apply)
Assumes independence between features
Often used for POS tagging and NER
Requires a large amount of training data
Provides probabilistic outputs
Cannot handle multi-class classification
Correct answers
Assumes independence between features
Often used for POS tagging and NER
Provides probabilistic outputs
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
In developing a sarcasm detection system for social media posts, which levels of NLP processing would be most crucial?
Phonological
Semantic
Discourse
Pragmatic
Syntactic
Correct answers
Semantic
Discourse
Pragmatic
In a Hidden Markov Model (HMM) for POS tagging, the transition probability from the tag NOUN (noun) to VERB (verb) is 0.3, and the transition probability from VERB (verb) to DET (determiner) is 0.4. What is the joint transition probability of the sequence NOUN → VERB → DET?
Correct answer: 0.12 (accepted within ±0.01)
In a Conditional Random Field (CRF) model for POS tagging, if there are 6 features per word and the sentence has 10 words, how many total feature values does the model compute for this sentence?
Correct answer: 60
In a POS tagging model, out of 500 total predictions made, 420 were correct. What is the accuracy of the model as a percentage (rounded to two decimal places)?
Correct answer: 84 (accepted within ±1)
In a binary classification task, the number of True Positives (TP) is 80, and the number of False Positives (FP) is 20. What is the precision of the model as a decimal (rounded to two decimal places)?
Correct answer: 0.8 (accepted within ±0.01)
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: 0.825 (accepted within ±0.025)
