Unsupervised technique for training an engagement classifier in chat-based group conversation

A method of enabling an engagement classifier for a group chatbot comprising:

• receiving a plurality of direct message conversations for respective users, by executing first instructions in a computer system;
• obtaining from each direct message conversation a collection of terms per user, by executing second instructions in the computer system;
• creating a knowledge representation for the users based on the collections of terms, by executing third instructions in the computer system;
• monitoring group messages to identify instances of user engagements, by executing fourth instructions in the computer system;
• training the engagement classifier by correlating the knowledge representations with the instances of user engagements, by executing fifth instructions in the computer system;
• creating a knowledge representation for the group chatbot based on another collection of terms derived from simulated conversations of the group chatbot;
• receiving real-time group chat different from the group messages;
• using the knowledge representation for the group chatbot with the engagement classifier to make a positive decision to engage in the group chat;
• and
• responsively engaging the group chatbot with the group chat.