Natural language understanding (NLU)
Natural language understanding (NLU) is the part of a conversational system that extracts meaning from free-form text or transcribed speech: intents, entities such as names, dates and addresses, and the relationships between them.
Also called: natural language processing, NLP, language understanding
NLU sits between hearing and deciding. Speech-to-text produces words; NLU produces structure: the caller wants an appointment (intent), on Thursday afternoon (a time entity), for the property on Elm Street (a location entity), and "the same as last time" refers to a service already on file.
Historically NLU was a separate component with hand-built grammars or trained classifiers for each intent and entity type. In agents built on large language models the understanding is folded into the model itself, which handles novel phrasing far better but makes its reasoning harder to inspect.
It is the reason a conversational agent can take "sometime next week, mornings are bad" and produce a usable booking request, where a form or a menu could not.