A user may say “I can’t log in” without mentioning a password reset. If a chatbot misses that goal, it may give a generic reply instead of helping. Intention is what the user wants to accomplish; an utterance is what they say. In an intent-based chatbot, a classifier groups utterances by meaning so the conversation can take an appropriate next step. section 1.1.3
Different words, same goal
“What time are you open?” and “What are your hours?” can share a store-hours intent and receive the same answer. “What cakes do you sell?” needs a different intent. Grouping equivalent requests saves the builder from writing a separate response for every phrasing. section 2.1.2
Recognizing an intent is a starting point, not necessarily a complete solution. A claim-status intent can start a process that gathers information, searches for a claim, and reports its status. A routing agent may instead identify the initial intent and pass the user to a specialist or human agent. section 10.1 section 1.1
Table 4.1 Example utterances may be handled differently based on the presence or absence of entities.
|
Utterance
|
Intent
|
Entity
|
Possible next step
|
|---|---|---|---|
|
“How many bags can I check?”
|
Bag_Allowance
|
Display bag check policy
|
|
|
“I want to book a flight”
|
Book_Flight
|
Collect destination
|
|
|
“I need a one-way ticket to Costa Rica”
|
Book_Flight
|
Costa Rica
|
Collect departure details
|
|
“I’d like to upgrade my seat to first class”
|
Flight_Upgrade
|
first class
|
Initiate upgrade process
|
Compare the flight-booking requests: both have the Book_Flight intent, but a stated destination can change what information the assistant needs to ask for next. section 4.1.3
Design and improve intents around actual demand:
- Start with paraphrases of a goal found in user interactions, then test whether the classifier recognizes new phrasings. A few obvious keywords may miss how people describe a problem. section 1.1.3 section 7.2
- Avoid adding every rare topic or splitting one useful goal into unnecessarily narrow intents. Use annotated logs and request volume to decide what belongs in the classifier; provide a fallback, search, or escalation path for uncovered requests. section 5.3.2 section 7.3.2
- Measure recognition with representative, previously unseen utterances, then check whether the answer or action actually advances the user’s goal. Improve one weak intent at a time and measure the effect. section 4.2.1 section 1.3.2
Intent classification offers control over predefined responses and flows, but its coverage is limited to the goals the system has been taught. Generative AI can help propose varied training utterances or handle some uncovered questions; its suggestions still need checking against real user demand and test data. section 2.1.2 section 5.3.2 section 7.3.2
