INVESTIGADORES
BENOTTI Luciana
congresos y reuniones científicas
Título:
Is conversational history used effectively in visual dialog systems?
Autor/es:
MAURICIO MAZUECOS; LUCIANA BENOTTI; ALBERTO TESTONI ; RAFFAELLA BERNARDI
Reunión:
Conferencia; Latin American Meeting In Artificial Intelligence (Khipu); 2019
Resumen:
Neural generative models for building dialog systems have gained popularity. But Sankar et al. (2019) has shown that these models do not use the dialog history effectively, for non visual dialog systems.In this poster we propose to explore what it takes for visual dialog systems to use dialog history effectively.We focus on question generation.