Paloma Martínez, F. Javier Calle, Dolores Cuadra, David del Valle, Jessica Rivero.
o F. J. Calle <fcalleinf.uc3m.es>
Interactor is an interaction platform based on Natural Interaction (human-like) techniques. It enables to implement a corpus-based Task Oriented Interaction Domain with little effort, and thus it assumes an application and provides access to it through Natural Interaction.
Once implanted onto a set of tasks, Interactor receives user interventions (represented through semantic structures) and handles the interaction in a human-like way. When required, it invokes the execution of some task(s), which are applications or drivers accessible from the server, and feeds the interaction back with its (their) results. Finally, it constructs system’s interventions (represented through semantic structures) and provides them. Besides, its Situation Model gathers knowledge on spatio-temporal features and objects within the interaction domain, and it is able to receive and process information about the user’s situation.
nteractor dialogue management is composed of four agents implemented in Java, running in Ecosystem, which provide basic services of agency registration, agent communication, and brokering, among others. The situation component is implemented trough two agents (also in Java) and based on a spatiotemporal database. The rest of the Interactor system (other agents) is also implemented in Java. All the mentioned databases are set on the Oracle TM 11g DBMS.
Interactor is currently running on a Sun Fire X4500 server, which also has the DBMS server, under Windows XP. However, due to the flexibility of the system architecture, the DBMS can be set on another server (also with different OS), and even each individual agent might run in different computers (with access to the DBMS server). Thus, system efficiency can be boosted as required.
For interaction performance, the following modules are required (1) Ecosystem (Multi-agent platform); (2) Interactor components; (3) Dialogue management agents; (4) Situation agents; (5) User model agent; and (6) Ontology agent. For following the internal processes and reasoning of the different agents, it is also required the (7) Tracer, a tool for monitoring the interaction. Finally, implementing new interaction domains involves corpus analysis tasks for which another tool is utilized, (8) the Cognos Toolkit (including relaxed grammars supported NLP and pragmatic analysis and annotation). This Toolkit is available for free use at the following URL: http://labda.inf.uc3m.es/doku.php?id=es:labda_lineas:cognos
Very few interaction systems count on a Situation Model (few prototypes, none commercial). This enriches interactive reasoning with the circumstantial aspect, apart from the situational services it can provide. In addition, this Situation Model is empowered by spatio-temporal database technology, ensuring versatility, scalability and efficiency.
Interactor and the Threads Model were the result of a PhD dissertation in 2005, and developed from the experience gained in several funded European and National research projects.
- Calle, J., Martínez, P., Valle, D., Cuadra, D. Towards the Achievement of Natural Interaction. Engineering the User Interface: from Research to Practice. © 2009 Springer (ISBN: 978-1-84800-135-0).
- Calle, J., García-Serrano, A., Martínez, P. Intentional Processing as a Key for Rational Behaviour through Natural Interaction. Interacting With Computers (ISSN: 0953-5438), Vol. 18/6, 1419—1446. © 2006 Elsevier.
- Cuadra, D., Calle, F.J., Rivero, J., Valle, D. Applying Spatio-Temporal Databases to Interaction Agents. International Symposium on Distributed Computing and Artificial Intelligence 2008 (DCAI’08). Volume: 50. Pag: 536-540. © 2009 Springer Berlin / Heidelberg (1615-3871).