The IL dataset has been produced by the CARESSES project (caressesrobot.org). The dataset is the collection of messages shared among the CARESSES components during interactions between the culturally competent robot and a person. Each IL file captures the events occurred during the encounter, the actions and status of the person (as perceived by the robot) and the actions of the robot, and it is acquired to allow offline analyses and replays of the events occurred during the interaction. Specifically, the IL dataset is organized as a collection of 3 automatically generated files per user: cahrimLog is the log of the "Culture-Aware Human-Robot Interaction Module" of the CARESSES architecture, and it reports: (i) all messages received and sent by this module from/to the other components of the architecture, (ii) the list of actions performed by the robot during the interaction together with (ii-a) their actual parameters, (ii-b) starting and (ii-c) ending time of each action and (ii-d) errors raised during its execution, if any; speechLog is the log of all dialogue instances between the robot and the person (with the person's utterances reported as understood by the robot); uAAL_log is the log of all messages exchanges by the components of the CARESSES architecture (CKB, where cultural knowledge is stored; CSPEM, responsible for planning the actions of the robot; CAHRIM, discussed above). IL files are encoded in CSV format, which is among the most readable formats for information storage. Each line corresponds to a record, i.e. all the info related to a message shared by any of the software components of the culturally competent robot during an encounter with a person. A record is divided into fields, separated by a delimiter (i.e., a semicolon in our case). The IL dataset, as a collection of quantitative data describing interactions between a person and an assistive robot can be useful to researchers aiming at defining guidelines, best practices and standards in the field of Human-Robot Interaction. Specific points of interest include: 1) identifying which robot actions are most frequently requested by people, and how, to design and develop more useful and easier to use assistive robots; 2) identifying recurring sequences of request/action/response that a robot or system for Ambient Assisted Living could rely on to exhibit predictive behaviours; 3) identifying preferred topics for conversation (and related phrasings), to design and develop more personal and adaptive assistive devices. Please notice also that, according to the original Data Management Plan, Interaction Logs are collected in two separate stages of the project: first in a laboratory setting in the course of Task 5.6 (m23 – m27) for benchmarking purposes, and then in the course of real-world experiments with care home residents in Task 6.3 (m28-m33) and Task 6.4 (m28-m33). However, we ultimately decided to publish only the Interaction Logs produced in the course of Task 5.6, performed in the last months of the project with international students enrolled in the Masters course on Robotics Engineering at University of Genova. The deviation from the original plan is necessary to ensure a proper quality of the Interaction Logs, that are now linked to video clips providing ground truth for comparison, recorded and published after seeking consent of the participant according to the provisions of the GDPR, and available upon request sending an email to info@caressesrobot.org
CARESSES Interaction Logs
2020-01-16
Research Data
Electronic Resource
English
DDC: | 629 |
British Library Online Contents | 2003
Taylor & Francis Verlag | 2016
|Online Contents | 2016
|