There are many existing chatbots based on decision tree technologies that require intense manual preparation and which can only cover a very limited number of scenarios. In addition, the results of this generation of assistants have generally been disappointing.
Chatbots using advanced and recent technologies, such as those developed for AVE-TAL, obtain much better results, though they are not yet widely available. While technological advances since 2018 have been considerable, as of now, they are rarely implemented.
The technologies developed for this project will make it possible to:
• exploit structured or unstructured knowledge;
• automatically generate questions;
• transfer learning so the tool is adaptable to several disciplines;
• use and optimize advanced open source NLP models;
• implement continuous learning to customize the behavior of the tool according to the learner’s profile;
• provide fluid assistance thanks to a human-machine interface that is augmented by conversational AI technology.
A well-trained assistant could therefore respond to 80% of user questions, continuously (24/7).