Natural language processing (NLP): how machines learned to handle words
Natural language processing, or NLP, covers the techniques that let a computer read, understand, translate and write human text or speech. If you use spell check, an online translator or an assistant like ChatGPT, you are already using NLP.
Data as of October 22, 2026
Key figures
| Level | Beginner |
|---|---|
| Origin | 1954 |
| Introduced by | Georgetown University et IBM (expérience de traduction automatique russe-anglais) |
What NLP means
Natural language processing is the branch of artificial intelligence that deals with words. Its goal is to let a machine work with human language, written or spoken. That covers a wide range of jobs: fixing a typo, translating a sentence, summarizing a report, flagging a message as spam, turning a voice into text or answering a question.
For decades, each of these tasks was handled separately, using rules written by linguists. Today most of them rely on machine learning: the system learns from examples rather than from hand-written rules.
Where you meet it every day
- At work: your email app suggests a quick reply (“Thanks, got it!”), sorts promotions into a separate tab and warns you about a suspicious message.
- On your phone: predictive text guesses your next word, and your voice assistant turns a spoken request into a command thanks to speech recognition.
- When traveling: a translation app helps you read a menu or message your host.
- With an AI assistant: when you ask a chatbot to rewrite an email, a large language model splits your text into small pieces and generates a reply one word at a time.
A short history
One of the first public demonstrations took place in January 1954: the Georgetown–IBM experiment, in which a computer automatically translated around sixty Russian sentences into English using only a handful of grammar rules and a small vocabulary (account of the experiment by historian John Hutchins). Over the following decades, researchers moved from linguistic rules to statistical methods. In 2017, the Transformer architecture paved the way for today’s language models.
Why it matters to you
Knowing that AI “processes” language without experiencing it the way we do helps you use it better. An NLP model is great at rewording, summarizing and translating, but it does not fact-check itself. To see how much text a model really handles, the token estimator shows how a sentence gets split up.
Common myths
- “NLP is new.” Research goes back to the 1950s. What is new is how good it has become for everyday users.
- “Machine translation is always reliable.” It usually gets the general meaning right, but can miss nuance, humor, or precise legal and medical wording.
- “If AI writes well, it understands everything.” Fluent text can still contain errors. Polished style is no guarantee of accuracy.
Related terms
- Large language model (LLM) · A program trained on huge amounts of text to predict what comes next in a text; it is the engine behind AI assistants.
- Chatbot · A program you talk to in writing or by voice, which replies in everyday language like a conversation partner.
- Speech recognition · Technology that turns spoken words into written text: dictation, captions, voice commands and meeting notes.
- Token · A small chunk of text (a word, part of a word or a symbol) that an AI model reads and writes; the unit for limits and pricing.
- Machine learning · A branch of AI where software learns patterns from examples instead of following rules written one by one by a person.
- Transformer · A neural network architecture built on attention, introduced in 2017, that underpins almost every large language model today.
Try it
Full glossary
Frequently asked questions
Is ChatGPT an example of NLP?
Yes. ChatGPT runs on a large language model, which is currently the most powerful NLP tool. But NLP is older and broader than chatbots: it also includes translation, speech recognition and spam filtering.
Does NLP actually understand language?
Not the way people do. Today's systems learn statistical patterns from huge amounts of text, which lets them respond in useful ways. They can still be confidently wrong, so it is worth checking anything important.
What is the difference between NLP and AI?
AI is the big umbrella. NLP is one branch of it, focused on language. Computer vision, which deals with images, is another branch.
Does NLP work equally well in every language?
No. English tends to get the best results because it dominates the training data. Widely written languages like Spanish or French work well, while languages with less online text are usually handled less accurately.
Data sources
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.