Navigating the complexities of Natural Language Processing (NLP) requires more than just reading textbooks; it demands a hands-on approach to how concepts are tested. Whether you are grappling with tokenization or trying to decode the intricacies of transformer models, practicing with actual exam questions is the most effective way to bridge the gap between theory and application.
Below, we have compiled a comprehensive Q&A guide based on common NLP examination themes. To help you prepare further, you can download the full PDF past paper via the link at the bottom of this page.
bellow is an exam paper download link
CDS-3400-NATURAL-LANGUAGE-PROCESSING-
above is the exam paper download link
Key Revision Questions & Answers
1. What is the primary difference between Stemming and Lemmatization?
While both techniques aim to reduce a word to its base form, their methods differ significantly. Stemming is a heuristic process that chops off the ends of words (e.g., “connection” becomes “connect”). It’s fast but often results in non-dictionary words. Lemmatization, however, uses a vocabulary and morphological analysis to return the “lemma” or dictionary form (e.g., “better” becomes “good”).
2. Explain the “Smoothing” technique in N-gram Language Models.
In N-gram models, you often encounter sequences of words in a test set that never appeared in the training data, resulting in a probability of zero. Smoothing (like Laplace or Good-Turing smoothing) reassigns some of the probability mass from seen events to unseen events. This prevents the model from crashing when it encounters a “zero-frequency” problem.
3. How does the “Attention Mechanism” solve the bottleneck issue in Encoder-Decoder models?
Traditional Encoder-Decoder architectures compress an entire input sequence into a single fixed-length vector. For long sentences, this “bottleneck” causes the model to forget earlier parts of the input. The Attention Mechanism allows the decoder to “look back” at all the hidden states of the encoder and focus on the most relevant parts of the input sentence for every word it generates.
4. Define “Named Entity Recognition” (NER) and its practical use.
NER is a subtask of information extraction that seeks to locate and classify entities in text into predefined categories such as names of persons, organizations, locations, and dates. It is widely used in news categorization, customer support ticket routing, and enhancing search engine efficiency.
5. What are the challenges associated with Ambiguity in NLP?
Ambiguity is perhaps the greatest hurdle in machine understanding. It occurs at multiple levels:
-
Lexical: A word having multiple meanings (e.g., “bank”).
-
Syntactic: A sentence having multiple tree structures (e.g., “I saw the man with the telescope”).
-
Semantic: The overall meaning is unclear despite knowing the words and structure.
Why Revise with Past Papers?
Reading notes gives you the “what,” but past papers give you the “how.” By simulating exam conditions, you identify which topics—like Hidden Markov Models or Word Embeddings—require more of your attention.
NLP is a rapidly evolving field. While the core mathematics of probability and linguistics remain stable, the application of deep learning grows every day. Revising past papers ensures you understand the foundational logic that modern AI tools like ChatGPT are built upon.
Download the Materials
Ready to test your knowledge? Use the link below to access the full document.
Last updated on: March 31, 2026