What Is Natural Language Processing? — lecture notes
How computers read, understand and generate human language — the tasks, the pipeline and how the field moved from rules to neural networks.
0:001. Introduction
Every search you type, every translation and every chatbot reply relies on natural language processing: the science of getting computers to work with human language.
0:112. NLP tasks
NLP covers many tasks. Classifying text, extracting names and facts, translating between languages, answering questions, and generating new text, from summaries to conversations.
0:223. The pipeline
Almost every NLP system follows the same path. Raw text is split into tokens. Tokens become vectors of numbers. A model processes those vectors, and produces an output: a label, a translation, an answer or new text.
0:384. Why it is hard
Language is hard for computers. Words are ambiguous: is a bank beside a river or full of money? And people use slang, typos, emojis, sarcasm and thousands of languages. Context is everything.
0:525. Three eras
NLP moved through three eras. First hand-written rules. Then statistics, counting words and word pairs. Then neural networks: word embeddings in 2013, the Transformer in 2017, and huge pre-trained models like BERT and GPT.
1:076. Recap
To recap. NLP classifies, extracts, translates and generates. Text becomes tokens and vectors before any model sees it. Ambiguity makes it hard. And the field moved from rules to statistics to neural networks.
Key takeaways
- NLP covers classification, extraction, translation, question answering and generation.
- Text is turned into tokens and then vectors before a model can process it.
- Ambiguity, context and variety make language difficult for computers.
- The field moved from rules to statistics to neural networks and pre-trained Transformers.
Check yourself
- What must happen before a neural model can process text?
Show answer
Text is converted into tokens and then numbers — Models compute with vectors, so text is tokenised and embedded.
- “I saw her duck” is hard for computers because of…
Show answer
Ambiguity — “Duck” could be an animal or an action.
- Which architecture (2017) underpins BERT and GPT?
Show answer
The Transformer — Both are built from Transformer blocks.
Go deeper
- Introduction to Natural Language Processing · The AI Lecture Hall
- Text Preprocessing: Tokenisation, Normalisation, Stemming and Lemmatisation · The AI Lecture Hall
© 2026 Janin A Apurba, CSE, AUST · Advanced ICT Officer, CNRS-UNHCR. All rights reserved. Notes for the animated lecture at https://ai-in-motion.vercel.app/watch/what-is-nlp.html