Home / Publications / Implementing a data mining approach to episodic memory modelling for artificial companions

Implementing a data mining approach to episodic memory modelling for artificial companions

TitleImplementing a data mining approach to episodic memory modelling for artificial companions
Publication TypeConference Paper
Year of Publication2011
AuthorsKeysermann MU, Freitas AA, Vargas PA
Refereed DesignationUnknown
Conference NameAISB 2011 Convention
Date Published04/2011
PublisherAISB
Conference LocationYork, UK
Keywordsbayesian classification, data mining, episodic memory, memory modelling
AbstractThe main goal of this work is to implement and test two different data mining approaches for retrieving and classifying the information within a computational episodic memory model developed for artificial companions. As the information stored in our episodic memory model reflects mainly certain past events we have elaborated an appropriate data structure and created the corresponding event data. Data mining techniques were then implemented for processing the knowledge within the memory model. The data mining task addressed here is classification and further prediction. Two Bayesian classifiers were evaluated by analysing prediction performance in general, comparing the results obtained in specific and more realistic scenarios. This work is a first step towards the full incorporation of data mining techniques to episodic memory modelling for artificial companions. Future work includes the processing of hierarchical data and other machine learning techniques in order to facilitate the creation of more believable artificial companions/robots.
URLhttp://dl.lirec.org/papers/KeysermannEtAl_HMAA2011.pdf
Posted by muk7 on Tuesday, 15 March, 2011 /

Photos

www.flickr.com