A
current trend in modern HCI is represented by Embodied Conversational
Agents (ECAs), even designed to run on the Web. They are virtual
3D human-like front ends coupled with software agents that are
able to engage in a conversation with a user and execute complex
tasks, such as, for example, searching for some specific information
or ordering some items from the catalogue of an online shop.
This paper presents SAMIR system, a framework to build intelligent
agents for the Web.
SAMIR consists of a 3D face which is animated to exploit expressions
which are perceived by the user; a custom version of the ALICE
chatterbot to chat with the user; and finally an XCS classifier
system to deal with the problem of keeping conversation and face
expressions coherent with each other. Experimental results, taken
from an online bookstore-based scenario, are presented at the
end.
