Modeling Lifecycle Overview

The complete lifecycle for managing your Conversational AI model includes:

  • Collecting Data: Gather utterances representative of what real users say
  • Measuring Accuracy: Measure the system to see how well it is performing and where it is having problems
  • Training The Model: Make changes to the model to address errors identified via testing and monitoring
  • Monitoring Interactions: Once the model is trained and put in production, it must be monitored on an ongoing basis to see how real users interact - figuring out what works and what does not.

Taken together, this is the lifecycle that defines building and optimizing a Conversational AI bot. This lifecycle is not a finite process, but instead is effective as long as the bot is live and interacting with real users.