Top 20 Tricky MCQs on RAG (Retrieval-Augmented Generation) with Answers | NLP & LLM Practice

✔ Scroll down and test yourself — answers are hidden under the “View Answer” button.

☰ Quick Links - Browse Related MCQs

🚨 Quiz Instructions:
Attempt all questions first.
✔️ Click SUBMIT at the end to unlock VIEW ANSWER buttons.

Quiz Mode:

Top 20 Tricky MCQs on Retrieval-Augmented Generation (RAG) with Answers

Retrieval-Augmented Generation (RAG) is a powerful technique that enhances large language models by combining information retrieval with text generation. It helps reduce hallucinations by grounding responses in external knowledge sources.

In this post, you will find 20 carefully designed MCQs on RAG with answers and explanations. These questions are useful for GATE, placements, NLP interviews, and data science exams.

1.

In a RAG system, what is the primary purpose of the retriever?

A. Generate natural language responses
B. Store training data
C. Fetch relevant documents from external knowledge
D. Fine-tune the language model

Correct Answer: C

Explanation:

The retriever fetches relevant external documents to provide context for generation.

2.

Which component converts text into vectors in RAG?

A. Decoder
B. Tokenizer
C. Embedding model
D. Optimizer

Correct Answer: C

Explanation:

Embedding models transform text into vector representations for similarity search.

An embedding model is a machine learning model that converts text (words, sentences, or documents) into numerical vectors in a high-dimensional space to enable semantic similarity search.

In the context of Retrieval-Augmented Generation (RAG):

• The query is converted into a vector
• Documents are also stored as vectors
• Then similarity (e.g., cosine similarity) is computed to find relevant documents

3.

Which similarity metric is most commonly used in vector search?

A. Euclidean distance
B. Manhattan distance
C. Cosine similarity
D. Hamming distance

Correct Answer: C

Explanation:

Cosine similarity measures angular similarity and is widely used for embeddings.

4.

What happens if chunk size is too large in RAG?

A. Improves retrieval precision
B. Reduces granularity of retrieved information
C. Eliminates need for embeddings
D. Guarantees better context

Correct Answer: B

Explanation:

Large chunks reduce retrieval granularity, making precise matching harder.

More info on chunk size and granularity:
In Retrieval-Augmented Generation, documents are split into chunks before being embedded and stored. Large chunk sizes reduce retrieval granularity, making it harder to extract precise and relevant information.

What does “granularity” mean?
High granularity → fine, precise pieces of information Low granularity → large, coarse blocks of text

What if each chunk size is too large?
If each chunk is very big, then:

• Each chunk contains too much mixed information
• Embeddings become less specific
• Retrieval returns broad but less relevant context
• This might result in the system cannot pinpoint the exact relevant passage.

That is exactly: Reduced granularity of retrieved information.

5.

Which problem does RAG primarily aim to mitigate?

A. Overfitting
B. Hallucination
C. Underfitting
D. Gradient vanishing

Correct Answer: B

Explanation:

RAG reduces hallucinations by grounding outputs in retrieved external knowledge.

Why do LLMs hallucinate?

Standard language models generate answers based on internal (parametric) knowledge. If they don’t know the answer, or have outdated/incomplete knowledge, they may confidently generate incorrect information (hallucination).

How Retrieval-Augmented Generation reduces hallucinations?

RAG changes the process from: “Generate from memory” to “Retrieve → then generate based on evidence”

Steps in RAG to avoid hallucination

1. Query comes in.
2. Retrieval: RAG searches documents/databases, fetches relevant, real information. This is called grounding.
3. Context injection: The retrieved content is added to the prompt.
4. Controlled generation: The model now relies on actual retrieved facts not just its internal memory.

Note: RAG does NOT eliminate hallucination completely. It only reduces hallucination and improves factual accuracy.

6.

In RAG, what is “top-k retrieval”?

A. Selecting k training samples
B. Retrieving k most similar documents
C. Training k models
D. Generating k outputs

Correct Answer: B

Explanation:

Top-k retrieval returns the most relevant documents based on similarity.

7.

Which database is typically used in RAG systems?

A. Relational DB
B. Graph DB
C. Vector DB
D. Key-value store

Correct Answer: C

Explanation:

Vector databases efficiently store and search embeddings.

Why a Vector DB is used in RAG?

In Retrieval-Augmented Generation, the goal is to find the most relevant information based on meaning, not just exact words. That’s where a vector database becomes essential. A vector database is used in RAG to store embeddings and perform fast semantic similarity search for retrieving relevant documents.

RAG works with embeddings (vectors), not raw text. So instead of “Find documents containing the exact keyword”, it does “Find documents that are semantically similar to the query”.

What a Vector DB actually does?

A vector database:

• Stores embeddings (numerical vectors of text)
• Performs similarity search (e.g., cosine similarity)
• Quickly retrieves top-k relevant documents

8.

What is the role of the generator in RAG?

A. Retrieve documents
B. Rank documents
C. Generate final response
D. Store embeddings

Correct Answer: C

Explanation:

The generator (LLM) produces the final response using retrieved context.

9.

Which factor most affects retrieval accuracy?

A. GPU memory
B. Embedding quality
C. Batch size
D. Learning rate

Correct Answer: B

Explanation:

High-quality embeddings improve semantic matching and retrieval performance.

10.

What is “chunk overlap”?

A. Duplicate training
B. Overlapping text segments
C. Parallel processing
D. Model ensembling

Correct Answer: B

Explanation:

Chunk overlap ensures that important context is not lost between adjacent chunks.

11.

Which architecture is commonly used in RAG generators?

A. CNN
B. RNN
C. Transformer
D. GAN

Correct Answer: C

Explanation:

Modern language models used in RAG are based on the Transformer architecture.

12.

Why is re-ranking used in RAG?

A. Reduce model size
B. Improve retrieval relevance
C. Speed up training
D. Compress embeddings

Correct Answer: B

Explanation:

Re-ranking refines the initially retrieved documents to improve relevance.

13.

What happens if irrelevant documents are retrieved in RAG?

A. Model crashes
B. Hallucination increases
C. Training stops
D. Accuracy improves

Correct Answer: B

Explanation:

Irrelevant context can mislead the generator and increase hallucinated outputs.

14.

Which technique helps reduce latency in RAG?

A. Increasing chunk size
B. Caching embeddings
C. Increasing top-k
D. Using larger models

Correct Answer: B

Explanation:

Caching avoids recomputation and speeds up repeated queries.

15.

What is “dense retrieval”?

A. Keyword-based search
B. Embedding-based search
C. Rule-based search
D. Index-based search

Correct Answer: B

Explanation:

Dense retrieval uses continuous vector embeddings for semantic search.

16.

Which is NOT a component of a standard RAG system?

A. Retriever
B. Generator
C. Knowledge source
D. Discriminator

Correct Answer: D

Explanation:

Discriminators are used in GANs, not in RAG architectures.

17.

Why is normalization applied to embeddings?

A. Reduce size
B. Improve similarity comparison
C. Speed training
D. Remove noise

Correct Answer: B

Explanation:

Normalization ensures consistent and meaningful similarity calculations.

18.

Which trade-off is critical in RAG systems?

A. Bias vs variance
B. Latency vs accuracy
C. Precision vs recall only
D. Memory vs CPU

Correct Answer: B

Explanation:

Increasing retrieval depth improves accuracy but also increases latency.

19.

What is hybrid retrieval?

A. Multiple generators
B. Combining dense and sparse retrieval
C. Using multiple GPUs
D. Ensemble learning

Correct Answer: B

Explanation:

Hybrid retrieval combines semantic (dense) and keyword (sparse) search.

20.

Which step must occur before similarity search in a RAG pipeline?

A. Generation
B. Retrieval
C. Query embedding
D. Re-ranking

⬆ Back to Top

Time Complexity of Machine Learning Algorithms Explained (With Infographic)

Time Complexity of Machine Learning Algorithms (Visual Guide)

Choosing the right machine learning algorithm is not just about accuracy—it also depends heavily on time complexity. Understanding how algorithms scale with data helps in building efficient and scalable models.

This infographic provides a clear comparison of the training and inference complexity of popular machine learning algorithms used in real-world applications.

Table of Contents 1. Understanding Complexity Terms 2. Linear Models 3. Tree-Based Models 4. SVM and KNN 5. Other Algorithms 6. Key Takeaways 7. When to use which algorithm?

1. Understanding Complexity Terms

The time complexity of machine learning algorithms depends on several important factors:

• n → Number of data samples
• m → Number of features
• c → Number of classes
• k → Number of clusters
• i → Number of iterations

Insight: As n and m increase, computational cost grows significantly.

2. Linear Models

Linear Regression and Logistic Regression are among the most efficient algorithms in machine learning.

They scale well with large datasets and are widely used in production systems.

Key Point: Fast training and inference → ideal for large-scale problems

3. Tree-Based Models

Decision Trees and Random Forests are powerful models but come with higher computational cost.

While Decision Trees are relatively fast, Random Forest increases complexity due to multiple trees.

Trade-off: Better accuracy but higher training time

4. SVM and KNN

Support Vector Machines (SVM) are computationally expensive, especially for large datasets due to quadratic and cubic complexity.

K-Nearest Neighbors (KNN) has almost no training cost but suffers from very slow inference since it compares with all data points.

Important: KNN is not suitable for real-time systems with large datasets

5. Other Algorithms

Naive Bayes is extremely fast and works well for text classification tasks.

Dimensionality reduction techniques like PCA and t-SNE are computationally expensive but useful for visualization and feature reduction.

K-Means clustering depends heavily on iterations and number of clusters.

Observation: Simpler models are faster, but complex models provide richer insights

6. Key Takeaways

Each algorithm has its own trade-offs between speed and performance.

Summary:
• Linear models (Linear/Logistic Regression) → scale well with features → efficient for large datasets → Fast & scalable
• Tree models → Decision trees (fast but can become slow in worst-case scenarios (O(n²))) → Random forest (increased training cost) → Balanced performance
• SVM → computationally expensive. not ideal for very large datasets → High accuracy but slow
• KNN → No training cost but slow inference → bad for real-time systems
• Naive Bayes → Fastest baseline model
• PCA & t-SNE → costly dimensionality reduction techniques
• K-Means → depends heavily on iterations and clusters

7. When to use which algorithm?

Each algorithm has its own trade-offs between speed and performance.

When to use ML algorithms?
• Use Linear/Logistic Regression for scalable and interpretable models.
• Use Random Forest when accuracy matters more than speed.
• Use SVM for small to medium datasets with high dimensions.
• Use KNN only when dataset is small.
• Use Naive Bayes for text classification problems.

Conclusion

Understanding time complexity helps you choose the right machine learning algorithm based on your dataset size and performance requirements.

In practice, selecting an algorithm involves balancing accuracy, speed, and scalability.

Infographic Credit:
This infographic is created by ANAS ALOOR and shared here for educational purposes with permission.
🔗 View Original Creator Profile

🏠

🏠 Go to Home

L1 vs L2 Regularization Explained Visually | Lasso vs Ridge Infographic

L1 vs L2 Regularization Explained (Visual Guide to Avoid Overfitting)

Regularization is a key technique in machine learning used to prevent overfitting and improve model generalization. Among the most widely used methods are L1 (Lasso) and L2 (Ridge) regularization.

This infographic provides a simple and intuitive explanation of how L1 and L2 regularization work, how they differ, and when to use each.

Table of Contents 1. What is Overfitting? 2. L1 Regularization (Lasso) 3. L2 Regularization (Ridge) 4. Key Differences 5. Quick Takeaway

1. What is Overfitting?

Overfitting occurs when a machine learning model memorizes training data instead of learning general patterns. As a result, the model performs well on training data but poorly on unseen data.

Problem: Poor generalization and unreliable predictions

2. L1 Regularization (Lasso)

L1 regularization adds a penalty equal to the absolute value of model weights:

Loss = Original Loss + λ × |w|

This method pushes some weights exactly to zero, effectively removing less important features from the model.

Key Benefit: Performs automatic feature selection by eliminating irrelevant features

3. L2 Regularization (Ridge)

L2 regularization adds a penalty equal to the square of model weights:

Loss = Original Loss + λ × w²

Instead of removing features, L2 reduces the magnitude of all weights, keeping every feature but lowering their influence.

Key Benefit: Keeps all features while reducing overfitting smoothly

4. Key Differences

L1 regularization creates sparse models by setting some weights to zero, while L2 regularization shrinks all weights evenly without eliminating features.

L1: Feature selection (sparse model)
L2: Smooth weight shrinkage (no feature removal)

5. Quick Takeaway

L1 regularization removes unnecessary features, whereas L2 regularization reduces their impact. Both techniques are essential for building robust and generalizable machine learning models.

Tip: Use L1 when feature selection is important and L2 when you want stable models with all features

Conclusion

Understanding the difference between L1 and L2 regularization is fundamental for improving model performance. Choosing the right technique depends on your dataset and problem requirements.

In practice, many modern models also use Elastic Net, which combines both L1 and L2 regularization for better performance.

Infographic Credit:
This infographic is created by HARIKARAN M and shared here for educational purposes with permission.
🔗 View Original Creator Profile

🏠

🏠 Go to Home

RAG Design Patterns Explained (2026 Guide) – Naive, Hybrid, Graph & Agentic RAG

RAG Design Patterns Explained (You Must Know in 2026)

Retrieval-Augmented Generation (RAG) is one of the most powerful techniques used in modern AI systems to improve accuracy, reduce hallucinations, and enable real-time knowledge retrieval. In this guide, we break down the most important RAG design patterns you must understand in 2026.

This infographic summarizes different architectures like Naive RAG, Hybrid RAG, Graph RAG, and Agentic RAG, helping you choose the right design for your applications.

Table of Contents 1. Naive RAG 2. Retrieve-and-Rerank 3. Multimodal RAG 4. Graph RAG 5. Hybrid RAG 6. Agentic RAG 7. Multi-Agent RAG

1. Naive RAG

Naive RAG is the simplest architecture where documents are split into chunks, embedded into vectors, and stored in a vector database. When a query is asked, relevant chunks are retrieved and passed to a generative model.

Use case: Basic QA systems, chatbots, document search

2. Retrieve-and-Rerank

This improves Naive RAG by introducing a reranking model. After retrieving candidate chunks, the system ranks them based on relevance before passing them to the LLM.

Advantage: Higher accuracy and better context selection

3. Multimodal RAG

Multimodal RAG extends retrieval to images, videos, and audio. It uses multimodal embeddings and models capable of understanding different data formats.

Use case: Medical imaging, video search, AI assistants

4. Graph RAG

Graph RAG integrates knowledge graphs to capture relationships between entities. Instead of simple vector similarity, it leverages structured connections for reasoning.

Best for: Complex reasoning, enterprise knowledge systems

5. Hybrid RAG

Hybrid RAG combines vector databases and graph databases, enabling both semantic similarity and structured reasoning.

Benefit: Balanced performance between accuracy and reasoning

6. Agentic RAG (Router-Based)

Agentic RAG uses AI agents to decide how to process queries. It can route queries to different tools, databases, or models dynamically.

Use case: Advanced AI assistants and enterprise copilots

7. Multi-Agent RAG

In this architecture, multiple agents collaborate to solve complex problems. Each agent specializes in tasks like retrieval, reasoning, or tool usage.

Future trend: Autonomous AI systems

Conclusion

RAG design patterns are rapidly evolving, moving from simple retrieval systems to complex agent-based architectures. Understanding these patterns helps you design scalable, accurate, and intelligent AI systems.

If you're building AI applications in 2026, mastering Hybrid and Agentic RAG architectures will give you a major advantage.

Infographic Credit:
This infographic is created by HARIKARAN M and shared here for educational purposes with permission.
🔗 View Original Creator Profile

Canonical Cover in DBMS MCQs with Answers (Top 10 Problems Explained)

✔ Scroll down and test yourself — answers are hidden under the “View Answer” button.

☰ Quick Links - Browse Related MCQs

🚨 Quiz Instructions:
Attempt all questions first.
✔️ Click SUBMIT at the end to unlock VIEW ANSWER buttons.

Quiz Mode:

Canonical Cover in DBMS – Top 10 MCQs with Answers and Explanations

Canonical Cover in DBMS is the minimal set of functional dependencies obtained by removing redundancy and extraneous attributes while preserving equivalence. It is widely used in normalization and database design.

Understanding Canonical Cover (Minimal Cover) is essential in Database Management Systems (DBMS), especially for normalization and functional dependency optimization. It helps simplify a set of functional dependencies by removing redundancy while preserving equivalence.

In this page, you will find top 10 carefully selected MCQs on Canonical Cover designed for university exams, GATE, and technical interviews. Each question is explained step-by-step to help you master key concepts like splitting dependencies, removing extraneous attributes, and eliminating redundant functional dependencies.

These practice questions will strengthen your problem-solving skills and help you quickly identify minimal covers in exam scenarios.

What is Canonical Cover in DBMS?

A Canonical Cover is a simplified set of functional dependencies that is equivalent to the original set but has:

• No redundant dependencies
• No extraneous attributes
• Single attribute on the right-hand side

It plays a crucial role in database normalization and helps in identifying candidate keys efficiently.

If you are new to canonical cover (minimal cover), I strongly suggest you to learn the required properties of every functional dependency to be in a canonical or minimal cover here.

1.

Given the functional dependency A → BC, what is the canonical cover?

A. A → BC
B. A → B, A → C
C. AB → C
D. A → B

Correct Answer: B

Explanation:

Canonical cover requires single attribute on RHS. So A → BC is split into A → B and A → C.

2.

Given FDs A → B, B → C, and A → C, what is the canonical cover?

A. A → B, B → C, A → C
B. A → B, B → C
C. A → C, B → C
D. A → B

Correct Answer: B

Explanation:

A → C is redundant since it can be derived using A → B and B → C.

3.

Given FDs AB → C and A → C, what is the canonical cover?

A. AB → C, A → C
B. AB → C
C. A → C
D. B → C

Correct Answer: C

Explanation:

AB → C is redundant because A alone determines C.

4.

Given FDs A → B, B → C, C → D, and A → D, what is the canonical cover?

A. A → B, B → C, C → D
B. A → B, B → C, C → D, A → D
C. A → D
D. A → B, C → D

Correct Answer: A

Explanation:

A → D is redundant since it can be derived through A → B → C → D.

5.

Given FDs AB → C and A → B, what is the canonical cover?

A. AB → C, A → B
B. A → B, A → C
C. A → C
D. AB → C

Correct Answer: B

Explanation:

Since A → B, AB reduces to A. Hence AB → C becomes A → C.

6.

Given FDs A → BC and B → C, what is the canonical cover?

A. A → B, A → C, B → C
B. A → B, B → C
C. A → BC
D. A → C

Correct Answer: B

Explanation:

Split A → BC into A → B and A → C. A → C is redundant since A → B and B → C imply it.

7.

Given FDs AB → C, C → D, and A → C, what is the canonical cover?

A. AB → C, C → D, A → C
B. A → C, C → D
C. AB → C, C → D
D. A → D

Correct Answer: B

Explanation:

AB → C is redundant because A alone determines C.

8.

Given FDs A → B and B → A, what is the canonical cover?

A. A → B, B → A
B. A → B
C. B → A
D. None

Correct Answer: A

Explanation:

Both dependencies are necessary; neither is redundant.

9.

Given FDs A → B, B → C, and AB → D, what is the canonical cover?

A. A → B, B → C, AB → D
B. A → B, B → C, A → D
C. A → B, B → C
D. A → D

Correct Answer: B

Explanation:

Since A → B, AB reduces to A. Thus AB → D becomes A → D.

10.

Given FDs A → BC, B → D, and A → D, what is the canonical cover?

A. A → B, A → C, B → D
B. A → BC, A → D
C. A → B, B → D
D. A → D

Correct Answer: A

Explanation:

A → BC splits into A → B and A → C. A → D is redundant since A → B and B → D imply it.

How to Become an AI Engineer in 2026: Complete Roadmap (Beginner to Advanced)

How to Become an AI Engineer in 2026: Complete Roadmap for Beginners

Want to become an AI Engineer in 2026? This practical roadmap shows you exactly what to learn—from Python fundamentals and Machine Learning basics to modern Generative AI tools like LLMs, RAG systems, and AI agents.

Whether you're a beginner or a developer transitioning into AI, this guide breaks down the essential skills, tools, and real-world projects you need to master to become a successful AI engineer.

AI Engineer Roadmap 2026: Step-by-Step Guide

If you want to become an AI engineer in 2026, you need a structured learning path that combines programming, machine learning, and modern generative AI tools. This roadmap breaks down everything you need to learn—from fundamentals to building real-world AI systems.

1. Foundations

The first step in your AI journey is building strong technical fundamentals. These skills form the base for everything you will learn later.

• Python programming
• Data Structures & Algorithms
• Working with APIs
• Git & Linux basics

Mastering Python and version control systems helps you write efficient, maintainable, and scalable code.

---

2. Machine Learning Basics

Once you have the fundamentals, the next step is understanding how machines learn from data.

• Supervised Learning
• Feature Engineering
• Model Training
• Model Evaluation

This stage teaches you how to build predictive models and evaluate their performance using real datasets.

---

3. Generative AI & LLMs

Generative AI is the most important skill for modern AI engineers. It focuses on working with large language models (LLMs) and intelligent systems.

• Prompt Engineering
• Embeddings
• Vector Databases
• RAG (Retrieval-Augmented Generation)

These concepts help you build AI applications like chatbots, knowledge assistants, and intelligent search systems.

---

4. AI Engineering Stack

To deploy real-world AI applications, you need to learn the modern AI engineering stack.

• FastAPI for backend APIs
• LangChain / LangGraph frameworks
• Vector Databases (pgvector, Pinecone)
• Docker & Cloud platforms

This stack enables you to build scalable, production-ready AI systems.

---

5. Build Real AI Systems

The final and most important step is applying your knowledge through hands-on projects.

• AI Chatbots
• AI Agents
• Document AI systems
• Automation workflows

Building projects not only strengthens your skills but also helps you create a strong portfolio to showcase your expertise.

Final Thoughts

The future AI engineer is not just a coder but a builder, architect, and problem solver. By following this roadmap and consistently building projects, you can successfully transition into AI engineering in 2026.

Infographic Credit:
This infographic is created by Brij kishore Pandey and published here with permission.
Original source: LinkedIn Profile

🏠

🏠 Go to Home

Top 20 Named Entity Recognition (NER) MCQs with Answers & Explanations | NLP Practice Questions

✔ Scroll down and test yourself — answers are hidden under the “View Answer” button.

☰ Quick Links - Browse Related MCQs

🚨 Quiz Instructions:
Attempt all questions first.
✔️ Click SUBMIT at the end to unlock VIEW ANSWER buttons.

Quiz Mode:

Top 20 Named Entity Recognition (NER) MCQs with Answers

What is Named Entity Recognition (NER) in NLP?

Named Entity Recognition (NER) is a core task in Natural Language Processing (NLP) that identifies and classifies important entities in text into predefined categories such as Person, Location, Organization, Date, Time, and Money. It helps machines understand real-world objects mentioned in unstructured text data.

For example, in the sentence "Elon Musk founded SpaceX in 2002", NER identifies:

• Elon Musk → Person
• SpaceX → Organization
• 2002 → Date

Why is Named Entity Recognition Important?

• Improves information extraction from text
• Used in chatbots and virtual assistants
• Enhances search engines and recommendation systems
• Supports resume parsing and document analysis
• Plays a key role in AI and data science applications

NER MCQs for Practice

Below are 20 carefully selected Named Entity Recognition MCQs with answers and explanations to help you prepare for exams, interviews, and competitive tests in NLP, Machine Learning, and Data Science.

1.

What is the primary goal of Named Entity Recognition (NER)?

A. Translate text
B. Identify and classify entities in text
C. Generate summaries
D. Perform sentiment analysis

Correct Answer: B

Explanation:

NER extracts entities such as names, locations, organizations, and dates from text and classifies them into predefined categories.

2.

Which of the following is NOT a typical NER entity category?

A. Person
B. Location
C. Verb
D. Organization

Correct Answer: C

Explanation:

NER identifies real-world entities, not grammatical categories like verbs.

3.

In the sentence “Apple released the new iPhone in California”, what is “Apple”?

A. Location
B. Organization
C. Person
D. Product

Correct Answer: B

Explanation:

“Apple” refers to a company (organization) rather than a fruit in this context.

4.

Which tagging format is commonly used in NER?

A. ASCII
B. BIO tagging
C. UTF-8
D. JSON

Correct Answer: B

Explanation:

BIO tagging marks the beginning, inside, and outside of named entities.

5.

What does the “B” in BIO tagging represent?

A. Boundary
B. Begin
C. Base
D. Block

Correct Answer: B

Explanation:

“B” indicates the beginning of a named entity.

6.

Which of the following is an example of BIO tagging?

A. B-PER, I-PER, O
B. PER, LOC, ORG
C. NN, VB, JJ
D. POS, NEG

Correct Answer: A

Explanation:

BIO tagging labels tokens based on their position within entities.

7.

Which algorithm is commonly used in traditional NER systems?

A. K-Means
B. Conditional Random Fields (CRF)
C. Apriori
D. PCA

Correct Answer: B

Explanation:

CRFs are widely used for sequence labeling tasks like NER.

8.

Why are CRFs preferred over HMMs in NER?

A. Faster training
B. No labeled data needed
C. Ability to model arbitrary features
D. Simpler implementation

Correct Answer: C

Explanation:

CRFs allow flexible feature engineering without strong independence assumptions.

9.

Which modern model achieves state-of-the-art performance in NER?

A. CNN only
B. LSTM
C. Transformer (BERT)
D. Decision Tree

Correct Answer: C

Explanation:

Transformers like BERT capture contextual meaning effectively for NER.

10.

What is the role of tokenization in NER?

A. Translate text
B. Split text into tokens
C. Remove stopwords
D. Generate embeddings

Correct Answer: B

Explanation:

Tokenization breaks text into smaller units like words or subwords, which are essential for NER processing.

⬆ Back to Top

11.

Which metric is commonly used to evaluate NER models?

A. Accuracy only
B. BLEU score
C. Precision, Recall, F1-score
D. RMSE

Correct Answer: C

Explanation:

F1-score balances precision and recall, making it the most suitable metric for NER evaluation.

⬆ Back to Top

12.

What challenge does NER face with ambiguous words like “Amazon”?

A. Overfitting
B. Polysemy
C. Tokenization error
D. Stemming

Correct Answer: B

Explanation:

Polysemy refers to words having multiple meanings depending on context, which is a major challenge in NER.

⬆ Back to Top

13.

Which dataset is widely used for NER benchmarking?

A. MNIST
B. ImageNet
C. CoNLL-2003
D. CIFAR-10

Correct Answer: C

Explanation:

CoNLL-2003 is a standard dataset used for evaluating NER systems.

⬆ Back to Top

14.

In BIO tagging, what does “O” represent?

A. Object
B. Outside entity
C. Operator
D. Output

Correct Answer: B

Explanation:

“O” indicates that the token does not belong to any named entity.

⬆ Back to Top

15.

Which is a limitation of rule-based NER systems?

A. Needs training data
B. Hard to scale and maintain
C. Too flexible
D. Requires deep learning

Correct Answer: B

Explanation:

Rule-based systems require manual updates and do not scale well to new domains.

⬆ Back to Top

16.

What is entity linking in NER?

A. Translating entities
B. Linking entities to a knowledge base
C. Removing entities
D. Tokenizing entities

Correct Answer: B

Explanation:

Entity linking maps recognized entities to structured databases like Wikipedia or knowledge graphs.

⬆ Back to Top

17.

Which feature is important in traditional NER?

A. Pixel intensity
B. Word shape (capitalization)
C. Image resolution
D. Audio signal

Correct Answer: B

Explanation:

Features like capitalization and word patterns are strong indicators of named entities.

⬆ Back to Top

18.

What issue arises when multiple words form a single entity like “New York City”?

A. Token splitting
B. Multi-word entity recognition
C. Overfitting
D. Noise reduction

Correct Answer: B

Explanation:

NER systems must correctly identify and group multiple tokens into a single entity.

⬆ Back to Top

19.

Which technique improves NER performance in deep learning models?

A. Random initialization
B. Pretrained embeddings (BERT)
C. Removing context
D. Feature elimination

Correct Answer: B

Explanation:

Pretrained models like BERT capture deep contextual representations, significantly improving NER performance.

⬆ Back to Top

20.

Which is a real-world application of NER?

A. Image compression
B. Speech synthesis
C. Resume parsing
D. Sorting algorithms

Correct Answer: C

Explanation:

NER is widely used in resume parsing to extract names, skills, organizations, and other structured information.

⬆ Back to Top