WorksheetsIntroduction to Natural Language Processing (NLP)
Total questions: 102
Worksheet time: 51mins
Which statement best defines Natural Language Processing (NLP)?
A field enabling computers to analyze human language
A database system for storing large numeric datasets
A method for compressing multimedia files for storage
A protocol for networking devices across the internet
Why is NLP important in computing?
It encrypts all data using language models
It increases screen resolution for sharper text
It replaces binary with analog signals in hardware
It converts human language into digital representations
Which is a listed feature of natural languages?
They are governed by syntax, lexicon, and semantics
They always have fixed meanings that never change
They are processed only through numerical vectors
They avoid redundancy to prevent multiple meanings
Which example illustrates autogenerated captions as an NLP application?
Audio enhancement that removes background noise only
Static subtitles manually typed by editors before release
Live speech converted to on‑screen text during videos
A slideshow presenting images with decorative font overlays
Voice assistants use NLP primarily to do what?
Understand spoken commands and execute tasks
Encrypt conversations end‑to‑end for security
Render 3D graphics for virtual environments
Balance battery usage across mobile sensors
Language translation in NLP is best described as
Changing audio volume based on speech intensity
Transforming pixels into words for screen readers
Rewriting code comments into machine instructions
Converting text or speech from one language to another
What does sentiment analysis aim to detect in text?
Whether content is positive, negative, or neutral
How many users have read the content online
Which author wrote the text and when
The most frequent nouns across all documents
A news website wants to sort articles into Food, Sports, and Politics. Which NLP tool fits this need?
Text classification to assign predefined categories
Named entity linking to external knowledge bases
Topic modeling to discover hidden latent themes
Dependency parsing to map grammatical relations
Keyword extraction is mainly used to
Measure reading speed of different users
Generate entire paragraphs from short prompts
Translate idioms between unrelated languages
Automatically find important words and expressions
Which statement about natural language redundancy is accurate?
Redundancy should be removed from all languages
It occurs only in informal speech, never in writing
It prevents listeners from understanding context
Multiple expressions can convey the same meaning
Which scenario best demonstrates applying text classification in a practical workflow?
Encrypting emails using public key certificates
Improving microphone quality with noise suppression
Compressing documents to reduce storage footprint
Routing customer emails into support, billing, or sales
A company tracks customer reviews to find emotions about its brand. Which NLP technique directly addresses this task?
Automatic speech recognition to transcribe audio
Machine translation to expand language coverage
Sentiment analysis to judge polarity of opinions
Keyword extraction to list frequent technical terms
Which stage of NLP divides input into structural units like tokens within sentences and paragraphs?
Lexical Analysis step of processing
Syntactic Analysis of grammar
Semantic Analysis for meaning
Pragmatic Analysis for context
During which NLP stage are grammatically incorrect constructions eliminated by checking relationships among words?
Discourse Integration stage
Semantic Analysis of sense
Syntactic Analysis or parsing
Lexical Analysis of tokens
At what point does NLP ensure that phrases such as "hot ice cream" are rejected as meaningless?
Semantic Analysis stage
Pragmatic Analysis end
Lexical Analysis at start
Syntactic Analysis phase
Which NLP stage checks intended meaning and context, often discarding literal meaning to match real-world use?
Lexical tokenization step
Semantic coherence check
Discourse relation check
Pragmatic interpretation stage
What is the purpose of Discourse Integration in NLP?
Builds token dictionary lists
Forms story by linking sentences
Maps words to lemma roots
Parses clauses into trees
Which statement best contrasts scriptbots and smart-bots?
Scriptbots follow fixed scripts; smart-bots learn from data
Scriptbots are flexible; smart-bots are limited
Scriptbots require coding; smart-bots avoid programming
Scriptbots need large databases; smart-bots use scripts
Which trait is typical of a scriptbot?
Wide functionality with learning
Works around a programmed script
Requires coding to adopt AI
Uses bigger databases directly
Which trait is typical of a smart-bot?
Limited functionality overall
No language processing skills
Mostly free and easy to add
Flexible and powerful design
Why is text normalization used before deeper NLP tasks?
To generate new training labels
To convert raw text to canonical form
To encrypt language for privacy
To translate into human speech
Which example illustrates normalization to a canonical form?
Expanding text into paragraphs
Replacing words with emojis
Turning “goooood” into “good”
Keeping all punctuation exactly
Which sequence lists common normalization steps in plausible order?
Parsing → discourse → pragmatic check
Lemmatization → translation → parsing
Segmentation → tokenization → stopword removal
Stemming → tokenization → segmentation
What does sentence segmentation accomplish in preprocessing?
Splits corpus into sentences
Maps words to vocabulary
Removes stop words only
Checks grammar relations
In comparing human and computer language processing, which statement is accurate?
Machines handle ambiguity easily
Humans store information discretely
Computers require numeric inputs
Humans skip errors during typing
Which chatbot capability most relies on Discourse Integration?
Understanding a word list alone
Maintaining coherence across turns
Removing special characters only
Converting text to lowercase
Which step converts a sentence into separate tokens such as words and numbers?
Tokenization process
Lemmatization step
TFIDF weighting
Stemming process
Which words are typically removed during preprocessing because they add little meaning to a sentence?
All adjectives and adverbs
Nouns like cat, tree
Stop words like and, the
Rare words like quark
Why is text often converted to a common case (usually lower case) during preprocessing?
To remove punctuation marks
To avoid case‑sensitive mismatches
To improve grammar rules
To increase sentence length
What does stemming primarily do to words?
Finds word synonyms
Strips affixes to root
Counts term frequencies
Predicts topic labels
How does lemmatization differ from stemming?
Removes punctuation only
Counts words per document
Produces meaningful lemmas
Creates longer tokens
In the example table, the lemma for studies is which word?
studied
studying
student
study
What does a Bag of Words representation primarily capture from documents?
Semantic role labels
Word order patterns
Syntactic parse trees
Word occurrence counts
Which is NOT a step in a typical Bag of Words pipeline?
Create dictionary vocabulary
Text normalization
Parse dependency trees
Build document vectors
Term Frequency (TF) in a document is best described as which measure?
Probability of a topic
Inverse of document length
Number of unique words
Count of word occurrences
Inverse Document Frequency (IDF) becomes small for which kind of word across a corpus?
Very common stop words
Words with many syllables
Extremely rare terms
Proper nouns only
Which formula correctly defines IDF for a word W?
Term count divided by document length
TF multiplied by log of term count
Total documents divided by documents with W
Documents with W divided by total documents
Which statement about TFIDF weighting is correct?
TFIDF ignores term frequency entirely
High TF and high IDF yield high score
High TF and low IDF yield highest score
Low TF and high IDF always zero score
Which practical task commonly uses TFIDF in NLP applications?
Speech synthesis voices
Image edge detection
Information retrieval ranking
Graph shortest path search
During stop word filtering, which of the following is likely to be removed from texts?
data and model
the and for
vector and matrix
random and sparse
Given two documents, a word appears 5 times in one and 0 times in the other; it appears in only one of 100 documents in the corpus. Which option best describes its TFIDF behavior?
Highest TFIDF in second document
High TFIDF in first document
Low TFIDF in first document
Equal TFIDF in both documents
Which statement best defines computer vision in the context of AI?
A field enabling machines to derive meaning from visual data
A tool for replacing all manual image editing tasks
A method for compressing photos for faster storage
A technique for increasing display screen resolution
How does computer vision primarily differ from classical image processing?
CV uses no machine learning, IP depends on deep models
CV is a subset of IP with narrower applications
CV only edits pixels, IP understands scene semantics
CV interprets content for decisions, IP enhances image quality
In the Emoji Scavenger Hunt example, what capability is demonstrated?
Rendering 3D graphics for virtual environments
Matching real objects to categories using camera input
Encrypting photos before sharing with social apps
Compressing video streams for low bandwidth
Which is a correct pairing in the CV vs Image Processing table?
CV example: object detection; IP example: rescaling brightness
CV example: tone mapping; IP example: handwriting recognition
CV main goal: enhance pixels; IP main goal: make predictions
CV scope: subset of IP; IP scope: superset of CV
Which application most directly uses facial features to verify identity?
Facial recognition in attendance systems
AR filters for playful face overlays
Product search by uploaded photo
Lane detection for autonomous cars
What is the role of CV in smart homes and cities for security?
Encrypt wireless traffic between sensors
Generate synthetic voices for intercom systems
Provide internet connectivity to household devices
Recognize authorized individuals and track movements
How do social media face filters operate in real time?
Translate spoken words into text captions live
Randomly place stickers without tracking features
Detect and map facial landmarks, then overlay graphics
Blur the background to reduce network usage
Which outcome best illustrates Google’s Search by Image feature?
Composes automatic captions from audio content
Converts text to speech for accessibility features
Encrypts uploaded images using secure hashing
Returns results by matching visual patterns and shapes
In retail, what does customer behavior tracking with CV aim to achieve?
Optimize store layouts by analyzing movement paths
Replace all human staff with autonomous robots
Predict national economic trends from sales
Prevent theft by encrypting video recordings
Which statement about inventory management with CV is accurate?
Images cannot indicate stock levels reliably
Barcodes eliminate the need for any cameras onsite
Cameras monitor shelves and trigger restock analysis
Manual counting is faster than automated detection
What CV tasks are essential for self-driving cars?
Rendering movies at a higher frame rate
Adjusting tire pressure automatically
Streaming music based on driver preferences
Detecting vehicles, pedestrians, signs, and lanes
Which example reflects CV use in autonomous driving systems?
Autopilot that detects nearby vehicles and maintains lanes
Manual cruise control that fixes speed without sensors
Navigation app that only computes shortest road distance
Dashboard theme that changes colors with time
How is CV applied in medical imaging to support diagnosis?
Analyzing scans to detect abnormalities using models
Watermarking images to protect hospital branding
Replacing all radiologists with non-visual tests
Compressing MRI files for quick email transfer
What combination enables the Google Translate camera feature?
RFID tags on signs provide multilingual codes
LiDAR maps depth and GPS chooses language
GANs generate fonts and NLP guesses grammar
OCR detects text and AR overlays translated words
Which CV task describes assigning a single label to an image?
Classification for single-object recognition
Reconstruction for building 3D surface models
Segmentation for pixel-level contour labeling
Tracking for temporal association of objects
What does classification + localization add beyond classification?
Identifies the object and its position in the image
Generates synthetic training images automatically
Predicts future object motion from past frames
Converts 2D images into volumetric 3D models
Which statement best distinguishes CV scope relative to image processing?
CV only handles videos while image processing handles images
CV is unrelated and excludes pixel-level operations
CV is narrower than image processing in purpose
CV is a superset that includes image processing tasks
Which scenario shows CV improving the shopper experience?
Cashier-less stores detecting items and charging accounts
Printing paper coupons at manual checkout counters
Only scanning barcodes with handheld devices
Posting price labels without any sensor feedback
Which statement best defines the pixel value in an 8-bit grayscale image?
It is a float from 0 to 1 for hue
It is an integer from 0 to 255 for intensity
It is three integers storing red and blue
It is a binary value indicating edges only
In a digital image, what does resolution primarily indicate?
The brightness range of the image
The number of kernels used in CNNs
The number of pixels composing an image
The color depth per channel only
Which correctly compares RGB and grayscale images?
RGB stores one value per pixel, grayscale stores three
Both store three channels but different ranges
Both store one channel with different bit depths
RGB stores three channel values, grayscale stores one
In an 8-bit byte image, what do values 0 and 255 typically represent?
0 is green, 255 is red
0 is black, 255 is white
0 is red, 255 is blue
0 is white, 255 is black
Which option lists common low-level visual features extracted from images?
Layers, epochs, losses
Edges, corners, blobs
Weights, biases, neurons
Windows, menus, icons
What is a kernel in the context of convolution on images?
A matrix that slides over pixels
A random seed for training
A color map for RGB display
A loss function for classification
Which best describes the convolution operation used in CNNs?
Element-wise addition with random noise
Element-wise multiplication with a sliding kernel
Sorting pixel values across channels
Averaging entire images into a scalar
What is the primary role of the ReLU layer in a CNN?
It normalizes features to mean zero only
It converts RGB to grayscale before pooling
It removes negative activations introducing non-linearity
It reduces image resolution overall
Which statement about pooling in CNNs is most accurate?
Pooling increases feature map size
Average pooling is identical to ReLU
Pooling replaces kernels with larger strides
Max pooling selects most prominent activations
What does a fully connected layer do after convolution and pooling?
Extracts edges using Sobel filters
Flattens features and assigns labels
Converts images into RGB channels
Performs instance segmentation directly
Which task requires assigning a label to each pixel of detected instances?
Image classification only
Object detection only
Instance segmentation
Feature extraction
A 1280×1024 image is described as having high resolution because it has:
Multiple color channels and depth
Wide dynamic range of intensities
Large kernels for convolution
Many pixels arranged in rows and columns
Which misconception about convolution is most plausible? Choose the correct statement.
Convolution always blurs regardless of kernel
Convolution sorts pixels by their coordinates
Convolution highlights patterns via learned kernels
Convolution ignores spatial neighborhoods entirely
If an edge-detection filter is applied to a building photo, the output feature map will mainly show:
Only the brightest windows
Random noise across all pixels
Boundaries outlining structures
Uniform regions with constant color
For multiple objects in an image, which pathway combines classification and localization?
Histogram equalization process
Clustering without labels
Object detection for multiple items
Classification only for single item
Which statement best describes the hierarchical relationship among AI, ML, and DL?
DL contains AI and ML within it
AI contains ML and DL within it
AI and ML are separate from DL entirely
ML contains AI and DL within it
Which option correctly contrasts rule-based and learning-based models?
Rule-based use fixed rules; learning-based adapt from data
Both require no data and only expert rules
Rule-based adapt from data; learning-based use fixed rules
Both adapt from data after manual labels
In the Venn diagram of AI, ML, and DL, which circle is the smallest?
Deep Learning circle
Artificial Intelligence circle
Machine Learning circle
All circles are equal size
Which example best fits a rule-based system?
Chatbot with predefined question–answer pairs
Spam filter that updates with new patterns
Neural network learning from large datasets
Image classifier trained on labeled data
What is the primary input difference in the learning-based block shown?
Unlabeled dataset is provided to model
Only rules are provided to model
Only answers are provided to model
No data or rules are provided
Supervised learning primarily relies on which type of data?
Synthetic data generated randomly
Noisy data without any labels
Unlabeled data without any tags
Labeled data provided for training
Which task is classification rather than regression?
Estimating tomorrow’s temperature
Determining if an email is spam
Forecasting a stock’s closing price
Predicting a house’s sale price
Which statement about deep learning from the notes is accurate?
DL is a subset of ML using neural networks
DL is broader than AI and includes ML
DL replaces ML and rule-based systems
DL uses rules instead of learning data
Which outcome is expected from a rule-based approach block?
Labels learned without any rules
Clusters discovered from unlabeled data
Answers produced from data plus rules
Continuous values without fixed rules
Which scenario best illustrates supervised learning in practice?
Classifying images as dog or cat
Tuning rules by human experts
Grouping customers without labels
Exploring data with no given targets
Which distinction between regression and classification is correct?
Both avoid using labeled training datasets
Both predict only continuous numeric outputs
Regression predicts classes; classification predicts numbers
Regression predicts continuous values; classification predicts classes
Why are learning-based models considered adaptive compared to rule-based models?
They modify themselves using patterns in data
They rely entirely on fixed human-written rules
They require no data and only heuristics
They output the same answers despite changes
Which statement best describes unsupervised learning?
Learns from labeled examples with targets provided
Learns by memorizing training data without generalizing
Learns on unlabeled data to find patterns and groups
Learns via rewards from interacting with environment
What is the main goal of clustering in unsupervised learning?
Reduce dimensionality using principal components
Map states to actions maximizing long-term returns
Group similar data points into meaningful clusters
Predict future numeric values from sequences
Which scenario illustrates reinforcement learning most accurately?
A model grouping shoppers by buying habits
A model predicting salaries from past increments
An agent learning to park via trial-and-error rewards
A model labeling emails using provided categories
In reinforcement learning, what provides learning signal to the agent?
Human annotations for each data sample
Random initialization of network weights
Manual feature engineering by experts
Positive and negative feedback as rewards
Which pair correctly matches learning type with typical data?
Association learning — labeled class targets
Supervised learning — unlabeled data points
Unsupervised learning — unlabeled data points
Reinforcement learning — labeled image datasets
Which best explains the difference between ANN and CNN?
ANN cannot learn features; CNN has fixed filters
ANN requires labels; CNN works only without labels
ANN performs clustering; CNN performs classification only
ANN models human neurons; CNN handles image features
What is the primary advantage of artificial neural networks highlighted here?
They eliminate the need for training data entirely
They automatically extract features from raw inputs
They guarantee perfect accuracy on large datasets
They require manual rules for each decision
Which layer in a neural network receives final processed information?
Output layer
Hidden layer
Input layer
Convolution layer
What happens in the hidden layers of a neural network?
Data is acquired without processing it
Core processing using weights and biases occurs
Labels are attached to each training example
Predictions are displayed to the end user
Which is a sub-category of unsupervised learning described in the material?
Regression model
Classification model
Clustering model
Temporal-difference model
Which machine learning type is best when pre-existing data is insufficient but interaction feedback exists?
Unsupervised learning
Semi-supervised learning
Supervised learning
Reinforcement learning
A supermarket wants to group customers by purchase patterns without prior labels. Which approach fits best?
Association under reinforcement learning
Clustering under unsupervised learning
Q-learning under supervised learning
Regression under supervised learning
Which statement about CNNs aligns with the content?
CNNs are unrelated to deep learning methods
CNNs cannot differentiate objects within images
CNNs use only fixed, untrained filters for edges
CNNs assign learnable weights to image features
Identify the correct sequence of layers in a simple feedforward neural network.
Input → Hidden → Output
Hidden → Input → Output
Input → Output → Hidden
Output → Hidden → Input
Which description matches association models in unsupervised learning?
Assign class labels using training examples
Predict continuous outcomes from numeric inputs
Discover co-occurrences among items bought together
Map states to actions using cumulative rewards
Which summary statement correctly matches ML model to purpose?
Supervised: discover new patterns; Reinforcement: train with labels
Unsupervised: discover patterns; Supervised: determine relationships via training
Reinforcement: memorize data; Unsupervised: maximize reward signals
Supervised: act in environments; Unsupervised: require labeled targets
