Reengineering a Domain-Independent Framework for Spoken Dialogue Systems

Software Engineering, Testing, and Quality Assurance for Natural Language Processing, pages 68–76, Columbus, Ohio, USA, June 2008. c⃝2008 Association for Computational Linguistics Reengineering a domain-independent framework for Spoken Dialogue Systems Filipe M. Martins, Ana Mendes, M´arcio Viveiros, Joana Paulo Pardal, Pedro Arez, Nuno J. Mamede and Jo˜ao Paulo Neto Spoken Language Systems Laboratory, L2F – INESC-ID Department of Computer Science and Engineering, Instituto Superior T´ecnico, Technical University of Lisbon R. Alves Redol, 9 - 2◦ – 1000-029 Lisboa, Portugal {fmfm,acbm,mviveiros,joana,pedro,njm,jpn}@l2f.inesc-id.pt http://www.l2f.inesc-id.pt Abstract Our work in this area started as a re- search project but when L2F joined TecnoVoz, a Portuguese national consortium including Academia and Industry partners, our focus shifted to real-time professional solutions. The integration of our domain-independent Spoken Dialogue System (SDS) framework into commercial products led to a major reengineering process. This paper describes the changes that the framework went through and that deeply af- fected its entire architecture. The communi- cation core was enhanced, the modules inter- faces were redefined for an easier integration, the SDS deployment process was optimized and the framework robustness was improved. The work was done according to software en- gineering guidelines and making use of design patterns. 1 Introduction Our SDS framework was created back 2000 (Mour˜ao et al., 2004), as the result ofin three graduation theses (Cassaca and Maia, 2002; Mour˜ao et al., 2002; Viveiros, 2004), one of which evolved into a masters thesis (Mour˜ao, 2005). The framework is highly inspired on the TRIPS architecture (Allen et al., 2000): it is a frame-based domain-independent framework that can be used to build domain-specific dialogue systems. Every domain is described by a frame, composed by domain slots that are filled with user requests. When a set of domain slots is filled, a service is executed. In order to do so, the dialogue system interacts with the user until enough information is provided. From the initial version of the framework two systems were created for two different domains: a bus ticket vending system, which provides an in- terface to access bus timetables; and a digital vir- tual butler named Ambr´osio that controls home de- vices, such as TVs (volume and channel), acclima- tization systems, and lights (switch on/off and in- tensity) through the X10 electrical protocol and the IrDA (Infrared Data Association) standard. Since 2003, Ambr´osio is publicly available in the “House of the Future”1, on the Portuguese Telecommunica- tions Museum2. As proof of concept, we have also built a pro- totype system that helps the user while performing some task. This was tested for the cooking domain and the automobile reparation domain. After the successful deployment of the mentioned systems, we began developing two new automatic telephone-based systems: a home banking system and a personal assistant. These are part of a project of the TecnoVoz3 consortium technology migration to enterprises. To answer to the challenges that the creation of those new systems brought to light, the focus of the framework shifted from academic issues to interactive use, real-time response and real users. Since …

Content truncated to 500 words