Preparing training data and initial bot training

• An automatic aggregation of the dialogue history into clusters by key words
• Marking up training dialogues and assigning topics to messages in clusters
• Composing an hierarchical classifier of the service topics

Using a visual designer of service scenarios

• Creating service scenarios by topics of customers’ requests
• Applying basic bot skills of switching to the relevant scenario after extracting all significant facts from customers’ phrases thanks to usage of ML/DL algorithms

A pre-set recognition of human behavior patterns

Recognizing phrases beyond the main topic, which may occur at any stage of scenarios.
Specifying correct replies to the forms of:

• Greeting
• Goodbye
• Consent
• Rejection
• Displeasure
• 12 more replies beyond the context of the dialogs

Monitoring service efficiency of bots

• Online mode: using a customized dashboard to control metrics of robotiс services
• Offline mode: building detailed reports on running bots by sessions and service scenarios

Managing service quality of bots

• Correcting dialogues, in which bots committed errors
• Additional training of bots based on corrected data