NLP once relied on hand-written grammars and statistical models; today large language models handle most tasks with a single system. Classic NLP tasks include sentiment analysis, named-entity recognition, translation and speech-to-text pipelines.
Many languages remain underserved because far less text exists to train on.