dc.description.abstract | Task-oriented dialogue systems are computer systems that aim to provide an interaction
indistinguishable from ordinary human conversation with the goal of completing user-
defined tasks. They are achieving this by analyzing the intents of users and choosing
respective responses. Recent studies show that by personalizing the conversations with
this systems one can positevely affect their perception and long-term acceptance.
Personalised social robots have been widely applied in different fields to provide assistance.
In this thesis we are working on development of a scientific conference assistant. The goal
of this assistant is to provide the conference participants with conference information and
inform about the activities for their spare time during conference. Moreover, to increase
the engagement with the robot our team has worked on personalizing the human-robot
interaction by means of face and name recognition.
To achieve this personalisation, first the name recognition ability of available physical
robot was improved, next by the concent of the participants their pictures were taken
and used for memorization of returning users. As acquiring the consent for personal data
storage is not an optimal solution, an alternative method for participants recognition
using QR Codes on their badges was developed and compared to pre-trained model in
terms of speed. Lastly, the personal details of each participant, as unviversity, country of
origin, was acquired prior to conference or during the conversation and used in dialogues.
The developed robot, called DAGFINN was displayed at two conferences happened this
year in Stavanger, where the first time installment did not involve personalization feature.
Hence, we conclude this thesis by discussing the influence of personalisation on dialogues
with the robot and participants satisfaction with developed social robot. | |