An utterance labeling system helps chatbot teams turn user messages into training data. A reviewer can add a misunderstood utterance to a set of acceptable answers and assign the appropriate intent label. section 12.6
The system sits between conversation design and machine-learning training: labeled utterances are passed to models, while message logs reveal where new labels or dialog branches may be needed. section 12.6
Correcting a chatbot misunderstanding
If a chatbot marks a student’s answer as incorrect because it misunderstood the utterance, add that answer to the possible correct responses and label it with the appropriate intent. section 12.6
Figure 12.6 Conversation design workflow
The conversation design workflow shows how labeling, model training, backend interaction, message logs, and ongoing design revisions connect. section 12.6
