Flashcards · Natural Language Processing · Free
Natural Language Processing flashcards, generated for you.
Example Natural Language Processing study cards to learn from right now — then generate a full set from your own notes (plus a practice quiz) and export to Quizlet or Anki. Free, no account needed.
Example Natural Language Processing flashcards
What is Natural Language Processing (NLP) at its core?
The field of AI that enables computers to understand, interpret, and generate human language in a meaningful and useful way—like teaching machines to read and communicate.
Why is human language hard for computers to understand compared to structured data?
Language is ambiguous (words have multiple meanings), context-dependent (same phrase means different things in different situations), and symbolic—computers naturally work with numbers and patterns, not inherent meaning.
What is tokenization and why is it the first step in NLP?
Splitting text into smaller units (words, subwords, characters) so computers can process them. It's necessary because raw text is just a long string—we need discrete, countable pieces to work with mathematically.
What problem does word embeddings solve in NLP?
Converts words into vectors (lists of numbers) so that computers can capture meaning and relationships—words with similar meanings get similar vectors, enabling mathematical operations on language.
Explain the intuition behind word2vec (like Word2Vec Skip-gram model).
Learn word vectors by predicting context words from a target word (or vice versa). The model figures out that words appearing in similar contexts should have similar representations—no manual labeling needed.
What is the key insight of the Transformer architecture that revolutionized NLP?
Use attention mechanisms to let every word directly 'look at' every other word in parallel, capturing long-range relationships efficiently. This replaces sequential processing and allows the model to focus on the most relevant parts of text.
What does the attention mechanism do mathematically?
Computes weights (attention scores) for each word pair using Query-Key dot products, then uses these weights to create a weighted sum of Value vectors. This lets the model learn what to focus on for each word.
Why are large language models (LLMs) trained on next-token prediction instead of classification tasks?
Next-token prediction is a self-supervised task (no labels needed, just text), uses massive amounts of unlabeled data efficiently, and naturally teaches the model language structure and world knowledge before fine-tuning.
What is the difference between training a model and fine-tuning a pre-trained model?
Training learns from scratch (slow, needs lots of data); fine-tuning starts with a pre-trained model's learned patterns and adjusts them slightly for a specific task (fast, needs less data—transfer learning).
How do prompt engineering and in-context learning work with LLMs?
LLMs learn from examples and instructions in the prompt (context window) without updating weights. Clear prompts, few examples, and specific task framing guide the model's completion behavior—a form of rapid adaptation.
Make your own Natural Language Processing study set
Flashcards for related topics
Studying Natural Language Processing to build with AI? MindloomHQ turns it into real skills — structured courses, agent projects, and certificates.
Explore MindloomHQ →