Previous-year questions, by module
Every question from the six Mumbai University papers (May 2023 to May 2026), sorted by syllabus topic. Repeats are merged into one entry; other wordings sit under “Also asked as”.
How the paper works
- 3 hours, 80 marks. Same layout in all six papers.
- Q1 is compulsory: any 4 of 5 short questions. In May 24 and Dec 25, Q1 had only 4 parts, all compulsory.
- Q2 to Q6: answer any 3. Each has two parts (a, b).
- Four of six papers have an HMM tagging numerical. Three repeat the same CYK parse.
- Hobbs and Centering anaphora resolution is in every paper.
Times asked, by module
Asked in 3 or more of the 6 papers
- 6×M5 · 5.4Explain anaphora resolution using the Hobbs algorithm and the Centering algorithm.
- 4×M2 · 2.4Explain the Porter Stemmer algorithm with its rules.
- 3×M2 · 2.1Explain the various preprocessing operations used in Natural Language Processing.
- 3×M2 · 2.4What is the output of morphological analysis for a regular verb, an irregular verb, a singular noun and a plural noun?
- 3×M3 · 3.3 · 3.4What are the limitations of the Hidden Markov Model (HMM) and the Maximum Entropy (MaxEnt) model for POS tagging?
- 3×M3 · 3.6For the given grammar, parse the statement “The man read this book” using the CYK (CKY) algorithm.
- 3×M4 · 4.6What is the knowledge (dictionary) based approach to WSD? Explain the Lesk algorithm with a suitable example.
- 3×M5 · 5.1Explain discourse reference resolution in detail.
- 3×M6 · 6.1Explain the machine translation approaches used in NLP.
Introduction to NLP
1.3Stages in NLP
-
2×Explain the various stages of Natural Language Processing.Also asked as
- Explain the different stages of NLP. Also explain a generic NLP system. May 25
1.4Ambiguities and their types
-
2×Explain the ambiguities at each level of Natural Language Processing, with examples.Also asked as
- What are the different levels of ambiguity in natural language, and how can they be resolved in NLP? May 26
-
1×Differentiate between syntactic ambiguity and lexical ambiguity.
1.5Challenges of NLP
-
1×Explain the challenges of Natural Language Processing.
1.6Applications of NLP
-
1×Explain the applications of Natural Language Processing.
1.1 history and 1.2 language, knowledge and grammar were never asked on their own.
Word Level Analysis
2.1Tokenization, stemming, lemmatization and preprocessing
-
3×Explain the various preprocessing operations used in Natural Language Processing.Also asked as
- Explain the pre-processing steps generally used in NLP. May 25
Preprocessing is not named in the syllabus. It sits under 2.1 (tokenization, stemming, lemmatization) and lab experiments 2 and 3 (stop words, filtration).
-
2×What is word level analysis? Explain the following terms with examples: tokenization, stemming, lemmatization.Also asked as
- Illustrate the concept of tokenization and stemming in Natural Language Processing. May 24
2.2Inflectional and derivational morphology
-
2×Explain inflectional and derivational morphology with an example.Also asked as
- Compare derivational and inflectional morphology. May 25
2.3Regular expressions
-
1×Explain the significance of regular expressions in NLP.
2.4Morphological models: FST, morphological parsing
-
3×What is the output of morphological analysis for a regular verb, an irregular verb, a singular noun and a plural noun?Also asked as
- Represent the output of morphological analysis for a regular verb, irregular verb, singular noun and plural noun. Also explain the role of FST in morphological parsing with an example. May 23
-
2×Design an FST for regular and plural nouns.
2.4Porter Stemmer algorithm
-
4×Explain the Porter Stemmer algorithm with its rules.Also asked as
- Illustrate the working of the Porter Stemmer algorithm. May 24
- Explain the Porter Stemming algorithm in detail. Dec 24
- Explain the steps of the Porter Stemmer algorithm with an example. May 26
2.5 · 2.8N-gram language models and perplexity
-
2×Explain the N-gram language model and its applications.Also asked as
- Explain the n-gram language model with formulas. What is the role of perplexity in evaluating an n-gram language model? Dec 25
-
1×Explain the concepts of bigram and n-gram with formulas. For the training corpus below, apply the bigram model and answer the two parts.<s> I am Sam </s> <s> Sam I am </s> <s> Sam I like </s> <s> Sam I do like </s> <s> do I like Sam </s>1. Most probable next word after:(a) <s> Sam … (b) <s> Sam I do … (c) <s> Sam I am Sam … (d) <s> do I like …2. Which sentence gets the higher probability?(e) <s> Sam I do like </s> (f) <s> Sam I am </s> (g) <s> I do like Sam I am </s>
-
1×Explain how the N-gram model is used in spelling correction.
Spelling correction with n-grams is the noisy channel idea from 2.10 (self-learning).
2.9Smoothing
-
1×Explain Laplace smoothing and Good-Turing discounting. Compare their effectiveness.
2.6 sensitivity to the training corpus and 2.7 unknown words were never asked.
Syntax Analysis
3.1POS tagging and the Penn Treebank tag set
-
1×Discuss the challenges in part-of-speech tagging.
-
1×Define POS tagging. Explain why it is a necessary pre-processing step for higher-level NLP tasks like Named Entity Recognition (NER). Using the Penn Treebank tagset, assign the appropriate POS tag to each word in the sentence below, and briefly explain the reasoning for the tags of the highlighted words.
“The fast train from Dadar will arrive at platform seven shortly.”
3.2Rule-based, stochastic and transformation-based tagging
-
1×Compare rule-based, stochastic and transformation-based POS tagging.
-
1×What is hybrid POS tagging?
Transformation-based (Brill) tagging is the usual “hybrid” answer: it learns rules from a tagged corpus, combining rule-based and stochastic ideas.
3.3HMM and Viterbi for POS tagging
-
1×Explain the Hidden Markov Model for POS tagging.
-
2×For the corpus below (S marks the start and E the end of a sentence), create the transition matrix and the emission probability matrix. Apply the Hidden Markov Model and do POS tagging for the given sentence.
<S> Martin Justin can watch Will <E> <S> Spot will watch Martin <E> <S> Will Justin spot Martin <E> <S> Martin will pat Spot <E> N (Noun): Martin, Justin, Will, Spot, PatM (Modal verb): can, willV (Verb): watch, spot, patSentence to tag: “Justin will spot Will”
-
1×Consider the following tagged corpus. Compute the emission and transition probabilities for a bigram HMM. Also decode the sentence below using the Viterbi algorithm.<s> a/DT dog/NN chases/V a/DT cat/NN </s> <s> the/DT dog/NN barks/V loudly/RB </s> <s> a/DT cat/NN runs/V fast/RB </s>
Sentence to decode: “The cat chases the dog.”
-
1×Consider the following tagged corpus. Compute the emission and transition probabilities for a bigram HMM. Also decode the sentence below using the Viterbi algorithm.<s> the/DT teacher/NN explains/V the/DT lesson/NN </s> <s> the/DT student/NN reads/V a/DT book/NN </s> <s> teacher/NN guides/V student/NN </s>
Sentence to decode: “The student reads the book”
3.3 · 3.4Limitations of HMM and MaxEnt
-
3×What are the limitations of the Hidden Markov Model (HMM) and the Maximum Entropy (MaxEnt) model for POS tagging?
3.4Maximum Entropy model and CRF
-
2×Explain the Maximum Entropy model for POS tagging.Also asked as
- Explain how Maximum Entropy is used for sequence labeling. Dec 24
-
2×Explain how Conditional Random Fields (CRF) are used for sequence labelling.Also asked as
- Demonstrate the concept of the Conditional Random Field in NLP. May 24
3.5Parsing with a CFG
-
1×Construct a parse tree for the sentence below using the given CFG rules.S → NP VPNP → Det Adj N | Det NVP → V | V NPDet → “the”Adj → “tall”N → “girl”V → “sings”
Sentence: “The tall girl sings.”
3.6Bottom-up parsers: CYK, PCFG, shift-reduce
-
3×For the given grammar, parse the statement “The man read this book” using the CYK (CKY) algorithm.S → NP VPDet → that | this | a | theS → Aux NP VPNoun → book | flight | meal | manS → VPVerb → book | include | readNP → Det NOMAux → doesNOM → NounNOM → Noun NOMVP → VerbVP → Verb NP
Same grammar and sentence in all three papers.
-
1×Explain the use of Probabilistic Context Free Grammar (PCFG) in NLP.
-
2×Explain the Shift Reduce parser in NLP with an example.
3.7 top-down parsers (Earley, predictive) were never asked.
Semantic Analysis
4.1Introduction, meaning representation
-
1×Describe semantic analysis in Natural Language Processing.
4.2Lexical semantics
-
2×Describe lexical semantic analysis using examples.Also asked as
- Demonstrate lexical semantic analysis using an example. May 24
4.3WordNet and BabelNet
-
1×Compare the structure, contents and utility of two major language dictionaries: WordNet and BabelNet.
4.4Relations among lexemes and their senses
-
1×Explain with suitable examples the following relationships between word meanings: homonymy, polysemy, synonymy, antonymy, hypernymy, hyponymy, meronymy.
4.6Word Sense Disambiguation
-
1×What is Word Sense Disambiguation (WSD)?
-
3×What is the knowledge (dictionary) based approach to WSD? Explain the Lesk algorithm with a suitable example.Also asked as
- What is WSD? Explain the dictionary based approach to WSD. May 23
- Explain the Lesk algorithm for Word Sense Disambiguation. May 24
-
2×Explain how the supervised learning approach can be applied to word sense disambiguation.Also asked as
- (i) Why is there a need for WSD? (ii) Explain the Naive Bayes supervised algorithm for WSD. May 25
-
2×What is WSD? Explain the semi-supervised (Yarowsky) algorithm for WSD.Also asked as
- Short note: semi-supervised method (Yarowsky) and unsupervised method (HyperLex). May 25
4.5 semantic ambiguity was never asked by that name (Dec 24 asked syntactic vs lexical ambiguity, listed under 1.4).
Pragmatic and Discourse Processing
5.1Reference resolution
-
3×Explain discourse reference resolution in detail.Also asked as
- Explain reference resolution in detail. Dec 24
- What is reference resolution in Natural Language Processing? May 26
5.2Reference phenomena
-
1×Illustrate the reference phenomena for solving the pronoun problem.
-
1×Explain three types of referents that complicate the reference resolution problem.
In Jurafsky & Martin these are inferrables, discontinuous sets and generics.
5.3Syntactic and semantic constraints on coreference
-
2×Explain coreference resolution in NLP. Explain the agreement constraints with examples.Also asked as
- Explain the following syntactic and semantic constraints on coreference: (1) number agreement, (2) person and case agreement. May 25
5.4Anaphora resolution: Hobbs and Centering
-
6×Explain anaphora resolution using the Hobbs algorithm and the Centering algorithm.Also asked as
- Describe in detail the Centering algorithm for reference resolution. May 23
- Compare and contrast Hobbs’ algorithm and Centering theory. Dec 24
- How is anaphora resolution performed with the Hobbs and Centering algorithms? May 25, May 26
Applications of NLP
6.1Machine translation
-
3×Explain the machine translation approaches used in NLP.Also asked as
- Demonstrate the working of machine translation systems. May 24
- Short note on (i) machine translation and (ii) question answering system. May 26
-
2×Explain rule-based machine translation systems.
6.1Text summarization
-
2×Explain text summarization in detail.Also asked as
- What is text summarization? Dec 25
6.1Information retrieval
-
1×Explain the information retrieval system.
-
1×Explain information retrieval versus information extraction systems.
6.1Question answering
-
2×Explain the question answering system with the algorithmic approach.Also asked as
- Short note on (i) machine translation and (ii) question answering system. May 26
Sentiment analysis was never asked.