NLU (Natural Language Understanding)
Qu'est-ce que NLU (Natural Language Understanding)?
Cette entrée n'est actuellement disponible qu'en anglais.
NLU (Natural Language Understanding) is the component of NLP that turns a person's free-form text or transcribed speech into structured meaning — identifying what the person wants (their intent) and pulling out the actual details (entities like dates, product names, or phone numbers) instead of matching input against a fixed keyword list. Where NLP covers both understanding and generating language, NLU is specifically the understanding step, and it powers intent detection in chatbots, IVR menus, and conversational AI agents.
More detail
An NLU step does two jobs on each input: classify the intent (the goal the person has, like booking an appointment or checking a balance) and extract entities (the concrete details inside the phrasing, like a date, order number, or card type). A dialog step can then act on that structured output instead of reinventing parsing per feature.
NLU spans both text and voice: for a phone call, an automatic-speech-recognition step transcribes the caller's words and the NLU step then parses the transcript into intent and entities — the same pipeline an inbound SMS, chat, or email flows through.
On Orbit, the AI agent framework's inbound parsing runs an LLM-based NLU pipeline across SMS, chat, email, and transcribed voice-call text, mapping free-form language onto structured intent and entities rather than a hand-tuned keyword classifier per channel.
Questions fréquentes
- What's the difference between NLU and NLP?
- NLP is the broader field covering both understanding and generating human language. NLU is narrower — specifically the understanding step that extracts intent and entities from what a person said; the generation step that produces a reply is a natural-language-generation (NLG) concern, one part of NLP that NLU doesn't perform.
- Why can't a fixed keyword matcher replace NLU?
- A keyword matcher only recognizes phrasing it was explicitly programmed for. NLU parses free-form language, so it maps 'when's my order arriving?' and 'status of my shipment?' onto the same intent — a keyword list that names every phrasing grows without bound and still fails on the one variant nobody wrote down.
- Does NLU work on voice calls?
- Yes — in a voice pipeline, an automatic-speech-recognition step transcribes the caller's speech and NLU then parses the transcript into intent and entities. The NLU layer itself is channel-agnostic: it operates on the resulting text, whether it came from ASR, SMS, chat, or email.
See also
Build it on Orbit
Voice, messaging, email, video, and AI agents on one platform and one pay-as-you-go bill. Start free — no credit card required.