The research needed for the project consists of 2 parts. Based on the state of the art which proposes generic conversational agent systems (DialogFlow, Rasa) integrating into chat solutions (Messenger, WhatsApp), it will be a question of coupling such tools with an Artificial Intelligence able to profile the readers thanks to the history of the dialogues and to suggest the articles according to the profiles. A machine learning approach based on active training of user models will be at the heart of (potentially patentable) innovation. The user profiles database will be exploited through clustering techniques in order to highlight global trends in the interests of readers and thus offer dashboards or predictive tools to drive the editorial lines. The research will have to integrate the temporal dimensions which are different according to the treated themes.
La Liberté, a Fribourg daily newspaper, will function as a test bed by providing its archives and will take part in the tests. The project is developed by Djebots, a company that implements preprogrammed bots. The project will be used for the press but the solution will be usable in other areas. The project is profitable from 2021.