A field of AI dedicated to enabling computers to understand, interpret, generate, and reason about human language — in text and speech form — powering applications from chatbots to translation systems.
In Depth
Natural Language Processing bridges the gap between human communication and machine computation. Language is the most natural way humans interact with the world — and one of the most complex inputs a machine must process. NLP covers the full pipeline from raw text or speech to meaningful representation: tokenization (splitting text into units), parsing (understanding grammatical structure), semantic analysis (extracting meaning), discourse analysis (understanding multi-sentence context), and generation (producing coherent output).
The history of NLP moved through rule-based systems (hand-crafted grammars), to statistical methods (n-gram language models, TF-IDF), to the current era of neural NLP dominated by pre-trained Transformers. Each transition dramatically expanded what was possible. BERT and GPT showed that a model pre-trained on vast text data could solve almost any NLP task with minimal task-specific adaptation — a paradigm shift that made NLP-powered applications more accurate and easier to build than ever before.
Modern NLP is deeply intertwined with the rise of Large Language Models. Today's LLMs perform nearly every classical NLP task — translation, summarization, question answering, sentiment analysis, named entity recognition — without the specialized architectures that were previously required. At the same time, LLMs introduce new NLP challenges: hallucination detection, factuality verification, multi-hop reasoning evaluation, and the assessment of emergent capabilities that classical NLP benchmarks weren't designed to measure.
NLP is how AI learns the language of humans — transforming the unstructured, ambiguous stream of words into structured information that machines can process, respond to, and generate with increasing fluency.