Last reviewed:
What is NLP? Natural language processing definition and examples
NLP (natural language processing) is the field of artificial intelligence that enables a machine to analyse, understand and produce human language, written or spoken. Large language models (LLMs) such as ChatGPT or Claude are now its most visible form.
France's data protection authority, the CNIL, defines natural language processing as a multidisciplinary field combining linguistics, computer science and artificial intelligence. In business, NLP is used to sort emails, extract data from a contract, analyse customer reviews, translate, summarise or run a chatbot. Beware of the homonym: in personal development, "NLP" also stands for neuro-linguistic programming, a communication method with no connection to AI. The field has gone through three ages. First, hand-written rules: in 1966, Joseph Weizenbaum's ELIZA program (MIT) simulated a conversation by spotting keywords and applying scripts. Then statistics: in 1990, an IBM team published an approach to machine translation that learns correspondences between languages from already translated texts, instead of grammar rules. Finally, neural networks, and above all the Transformer architecture, introduced in June 2017 by a Google team in the paper "Attention Is All You Need". Today's LLMs all descend from it. This last age changed the economics of NLP. Previously, each task needed its own specialised model (one for classification, one for extraction, one for translation), trained on thousands of annotated examples. A general-purpose LLM handles most of these tasks from a simple instruction. Classic NLP keeps some advantages: a small specialised model is cheaper to run, faster, and easier to control on a narrow, repetitive task.
Concrete example
Illustrative case: a regional French health insurer with 400 employees receives 6,000 letters and emails a week (reimbursement requests, complaints, address changes, cancellations). Until 2023, a classic NLP tool, trained on examples annotated by staff, sorted them into twelve categories. Each new category took several weeks of annotation. In 2025, the insurer tests an LLM: it classifies messages, extracts the member number and summarises the request in two lines, from an instruction written in a single day. The insurer keeps the old classifier for bulk sorting, which is cheaper per message, and uses the LLM for long letters and complaints, where fine understanding of the text saves the most time.
Comparison
| Rule-based NLP | Statistical NLP | Transformers and LLMs | |
|---|---|---|---|
| Period | 1950s to 1980s | 1990s to 2010s | Since 2017 |
| Principle | Rules and dictionaries written by linguists | Probabilities learned from large corpora | Neural network trained on huge volumes of text |
| Example | ELIZA (1966) | IBM statistical translation (1990) | ChatGPT, Claude, Mistral |
| Strength | Predictable, explainable | Robust on a well-defined task | Versatile, steered by an instruction |
| Limit | Brittle outside its rules | One model and annotated data per task | Cost, hallucinations, harder to control |
FAQ
What is NLP in artificial intelligence?
It is natural language processing: the set of techniques that enable a computer to analyse, understand and produce text or speech. Machine translation, spell checkers, voice assistants, chatbots and LLMs all belong to NLP.
What is the difference between NLP (natural language processing) and neuro-linguistic programming?
They are unrelated. Natural language processing is a discipline of computer science and AI. Neuro-linguistic programming, also abbreviated NLP, is a communication and personal development method.
What is the difference between NLP and an LLM?
NLP is the field; an LLM is a type of model. An LLM is a large Transformer neural network trained on huge volumes of text, which performs most NLP tasks from an instruction. NLP existed long before LLMs, first with rules and then with statistical methods.
What are examples of NLP in business?
Automatic sorting and routing of emails, data extraction from invoices or contracts, sentiment analysis of customer reviews, summarising meeting notes, translating documentation, support chatbots, search across an internal document base.
What are the main stages in the history of NLP?
Hand-written rules (ELIZA, 1966), then statistical methods that learn from corpora (IBM's machine translation, 1990), then neural networks and the Transformer architecture (2017), on which today's LLMs are built.
See also
Further reading
Attention Is All You Need, Vaswani et al., arXiv, June 2017 (founding paper of the Transformer)
Sources
- Traitement automatique du langage naturel (natural language processing or NLP), definition, CNIL. https://www.cnil.fr/fr/definition/traitement-automatique-du-langage-naturel-natural-language-processing-ou-nlp
- ELIZA, a computer program for the study of natural language communication between man and machine, Joseph Weizenbaum, Communications of the ACM, vol. 9, no. 1, January 1966. https://dl.acm.org/doi/10.1145/365153.365168
- A Statistical Approach to Machine Translation, Brown et al. (IBM), Computational Linguistics, vol. 16, no. 2, 1990. https://aclanthology.org/J90-2002/
- Attention Is All You Need, Vaswani et al., arXiv, June 12, 2017. https://arxiv.org/abs/1706.03762