Spotlight: Learning, optimizing, or handcrafting chatbot components

14:00-14:10, January 28 @ 5ABC

Talk/ Overview

The lessons we learned from the design of a question-answering chatbot show that fully end-to-end training of the system is not possible due to the lack of data.  Therefore, while certain components can be trained end-to-end by adapting existing datasets, others must rely on handcrafted algorithms, and the entire system can only be optimized through repeated evaluations.

Talk/ Speakers

Andrei Popescu-Belis

Professor, HEIG-VD

AMLD / Global partners