If you find the content useful and wish to support our platform’s development, you can contribute any amount toward our production costs. Scan the UPI QR code for payment within India. Or use the Ko-fi link to process a secure payment via PayPal.
LLMs & NLP
25 Terms
A–Z Linked
LLMs & NLP
Concepts related to large language models, natural language processing, and language understanding.
Select a term below or scroll through the full category.
- Context
- Corpus
- Decoder
- Dialogue System
- Encoder
- Information Extraction
- Instruction Tuning
- Intent
- Language Model
- Large Language Model
- Named Entity Recognition
- Natural Language Generation
- Natural Language Understanding
- Open-weight AI
- Perplexity
- Question Answering
- Reinforcement Learning from Human Feedback
- Semantic Search
- Sentiment Analysis
- Sequence
- Similarity
- Text Classification
- Tokenisation
- Vector
- Vocabulary
Context
📘 Definition
Context is the surrounding information that helps an AI interpret the meaning of text or a conversation.
🪄 Simple Explanation
Context helps AI understand what you really mean.
🔑 Why It Matters
Providing context improves response quality.
💡 Example
Mentioning that an email is for a job application changes the AI's writing style.
See Also
Corpus
📘 Definition
A corpus is a large organised collection of text used to train or evaluate language models.
🪄 Simple Explanation
A corpus provides language examples for AI.
🔑 Why It Matters
High-quality corpora improve model performance.
💡 Example
Millions of books form a training corpus.
See Also
Decoder
📘 Definition
A decoder generates output tokens one at a time based on previous context.
🪄 Simple Explanation
The decoder produces the AI's response.
🔑 Why It Matters
Many language models use decoder architectures.
💡 Example
A decoder writes one word after another.
See Also
Dialogue System
📘 Definition
A dialogue system manages conversations between people and AI systems.
🪄 Simple Explanation
Dialogue systems support interactive communication.
🔑 Why It Matters
They underpin customer service and digital assistants.
💡 Example
A banking chatbot guides customers through account enquiries.
See Also
Encoder
📘 Definition
An encoder converts input text into meaningful internal representations that AI models can process.
🪄 Simple Explanation
An encoder helps AI understand input.
🔑 Why It Matters
Encoders are widely used in language understanding tasks.
💡 Example
A search model encodes documents before comparison.
Information Extraction
📘 Definition
Information extraction automatically identifies structured facts from unstructured text.
🪄 Simple Explanation
It turns text into structured information.
🔑 Why It Matters
It supports search and analytics.
💡 Example
Software extracts invoice numbers from scanned documents.
Instruction Tuning
📘 Definition
Instruction tuning is the process of further training a pretrained language model on examples of instructions and appropriate responses so that it becomes better at following user requests.
🪄 Simple Explanation
Instruction tuning teaches a language model how to respond more effectively to human instructions.
🔑 Why It Matters
Pre-training teaches broad language patterns, while instruction tuning helps transform a general language model into an assistant that can respond usefully to diverse tasks.
💡 Example
A model may be trained on examples pairing instructions such as 'Summarise this passage' with high-quality summaries.
See Also
Continue Learning
Intent
📘 Definition
Intent is the underlying purpose or goal behind a user's message.
🪄 Simple Explanation
Intent explains what the user wants to achieve.
🔑 Why It Matters
Intent detection is essential for chatbots and virtual assistants.
💡 Example
A support bot recognises that 'I can't log in' is a request for technical help.
See Also
Language Model
📘 Definition
A language model is an AI system trained to predict and generate human language by learning patterns from large collections of text.
🪄 Simple Explanation
A language model learns how words and sentences fit together.
🔑 Why It Matters
Language models are the foundation of modern conversational AI.
💡 Example
An AI assistant uses a language model to answer questions.
Large Language Model
📘 Definition
A Large Language Model (LLM) is an artificial intelligence model trained on vast amounts of text to understand, generate, summarise, translate, and respond to human language. Modern LLMs are typically based on transformer architectures and learn statistical patterns in language rather than memorising facts or following fixed rules.
🪄 Simple Explanation
A Large Language Model is an AI system that learns from enormous collections of text so it can understand and generate human language.
🔑 Why It Matters
Large Language Models power many of today's AI applications, including ChatGPT, Claude, Gemini, and Microsoft Copilot. Understanding LLMs provides the foundation for learning prompt engineering, conversational AI, and modern generative AI.
💡 Example
When you ask an AI assistant to explain a concept, write an email, summarise a report, or generate computer code, the response is typically produced by a Large Language Model.
Named Entity Recognition
📘 Definition
Named Entity Recognition (NER) identifies people, organisations, places, dates, and other entities within text.
🪄 Simple Explanation
NER finds important names in text.
🔑 Why It Matters
It helps organise and analyse information.
💡 Example
An AI extracts company names from news articles.
See Also
Natural Language Generation
📘 Definition
Natural Language Generation (NLG) is the process of producing human-like text from structured information or learned patterns.
🪄 Simple Explanation
NLG allows AI to write coherent text.
🔑 Why It Matters
It powers reports, summaries, and chatbots.
💡 Example
An AI writes a project summary from raw data.
Natural Language Understanding
📘 Definition
Natural Language Understanding (NLU) enables AI systems to interpret the meaning, intent, and context of human language.
🪄 Simple Explanation
NLU helps AI understand what people mean.
🔑 Why It Matters
It underpins conversational AI.
💡 Example
A chatbot identifies that a customer wants a refund.
Open-weight AI
📘 Definition
Open-weight AI refers to artificial intelligence models whose trained model weights are publicly available for others to download, run, fine-tune, or deploy. While the weights are accessible, the model may not be fully open source because its training data, code, or licence may still have restrictions.
🪄 Simple Explanation
Open-weight AI models let people use and customise the model because the trained weights are publicly available.
🔑 Why It Matters
Open-weight models have accelerated AI research, education, and commercial innovation by allowing developers and researchers to run powerful models on their own hardware, fine-tune them for specialised tasks, and build new applications without relying entirely on cloud-based proprietary services. However, having open weights does not necessarily mean the model is fully open source.
💡 Example
A university downloads an open-weight language model and fine-tunes it on medical research papers to create a specialised assistant for healthcare researchers.
See Also
- Large Language Model
- Foundation Model
- Model
- Inference
- Fine-Tuning
- Open Source AI
Continue Learning
Perplexity
📘 Definition
Perplexity is a metric used to evaluate how well a language model predicts a sequence of tokens. Lower perplexity generally indicates that the model assigns higher probability to the observed text.
🪄 Simple Explanation
Perplexity indicates how surprised a language model is by the text it is trying to predict.
🔑 Why It Matters
It is useful for comparing similar language models or training checkpoints, although it does not by itself measure factual accuracy, reasoning, safety, or usefulness.
💡 Example
When two similar language models are tested on the same corpus, the one with lower perplexity predicts the text more confidently on average.
See Also
Continue Learning
Question Answering
📘 Definition
Question answering enables AI to answer questions using learned knowledge or retrieved information.
🪄 Simple Explanation
AI answers questions in natural language.
🔑 Why It Matters
It powers many virtual assistants.
💡 Example
A student asks AI to explain Newton's laws.
Reinforcement Learning from Human Feedback
📘 Definition
Reinforcement Learning from Human Feedback (RLHF) is a training approach that uses human preferences or preference models to help shape an AI system's behaviour towards responses people judge to be more useful or appropriate.
🪄 Simple Explanation
RLHF uses human preferences to help an AI model learn which kinds of responses people prefer.
🔑 Why It Matters
RLHF has been used to improve instruction following and align language-model behaviour more closely with human expectations, although it does not guarantee correctness or safety.
💡 Example
Human reviewers may compare two model responses, and those preference signals can be used to train the system towards more desirable behaviour.
Semantic Search
📘 Definition
Semantic search retrieves information based on meaning rather than exact keyword matches.
🪄 Simple Explanation
Semantic search understands intent.
🔑 Why It Matters
It produces more relevant search results.
💡 Example
Searching for 'car' also finds documents about automobiles.
See Also
Sentiment Analysis
📘 Definition
Sentiment analysis determines whether text expresses positive, negative, or neutral opinions.
🪄 Simple Explanation
It measures the emotional tone of text.
🔑 Why It Matters
Businesses use it to understand customer feedback.
💡 Example
Reviews are classified as positive or negative.
See Also
- Natural Language Processing
- Opinion Mining
- Text Classification
Sequence
📘 Definition
A sequence is an ordered series of tokens or data processed by a language model.
🪄 Simple Explanation
Language models analyse sequences of tokens.
🔑 Why It Matters
Order is important in language understanding.
💡 Example
Changing word order changes sentence meaning.
See Also
Similarity
📘 Definition
Similarity measures how closely two pieces of information are related.
🪄 Simple Explanation
Similarity helps AI compare meaning.
🔑 Why It Matters
It improves recommendation and search systems.
💡 Example
An AI finds articles similar to a research paper.
See Also
Text Classification
📘 Definition
Text classification assigns documents or messages to predefined categories.
🪄 Simple Explanation
It automatically labels text.
🔑 Why It Matters
It supports spam detection and document organisation.
💡 Example
Emails are classified as spam or legitimate.
Tokenisation
📘 Definition
Tokenisation is the process of splitting text into tokens before it is processed by a language model.
🪄 Simple Explanation
Tokenisation prepares text for AI.
🔑 Why It Matters
It is the first stage of language processing.
💡 Example
A sentence is divided into tokens before inference.
See Also
Vector
📘 Definition
A vector is a numerical representation of information used by AI models to compare meaning mathematically.
🪄 Simple Explanation
Vectors convert information into numbers.
🔑 Why It Matters
Vectors enable semantic similarity.
💡 Example
Similar sentences have nearby vectors.
See Also
Vocabulary
📘 Definition
A model's vocabulary is the collection of tokens it recognises during language processing.
🪄 Simple Explanation
Vocabulary defines what a model can represent.
🔑 Why It Matters
It influences tokenisation and efficiency.
💡 Example
Frequently used words often become single tokens.