Natural language processing, or NLP, covers any system that works with human language — understanding it, generating it, or translating between languages. Before models could handle language well, computers needed a way to represent words as numbers, since that's the only thing neural networks understand.
Early approaches represented words as simple counts or one-hot lists, which missed meaning entirely — "car" and "automobile" looked as unrelated as "car" and "banana." A major breakthrough came from representing words as points in a mathematical space, where words used in similar contexts end up near each other, capturing meaning far more naturally.
Today's NLP systems, built on top of large language models, handle translation, summarization, sentiment analysis, and conversation with a fluency that would have seemed implausible a decade ago — though they still stumble on sarcasm, cultural nuance, and genuinely novel wordplay.