Self-Learning Virtual Agents for Contact Centers

The ODISAE project is designed to achieve a semantic parser of conversations between agents and customers online (via chat, email or on a forum) and use it to enrich a CRM system of non-existent semantic features in the systems currently available on the market.


This innovation will be used to describe conversations with several properties: status (is this conversation a success or a failure?), tone ((un) satisfaction, aggressiveness, fun,…), themes (themes covered in the conversation), structure (evolution of the conversation over time).

Learning Avatar

Based on this analysis of the interactions between client and agent, ODISAE will allow to: automatically updating the FAQ, detection experts, assess the quality of the answers made to customers by the agents, users or automatic systems, (crucial for the formation of agents) trigger actions over the interaction to help for sale or on the contrary prevent attrition, etc., analyze the use of different channels of interaction behaviour and in particular intermodality.

This post is also available in: French