DISC

Spoken Language Dialogue Systems and Components Best practice in development and evaluation

Esprit Long Term Research Concerted Action no. 24823

Action summary

Software engineering best practice forms a backbone for the training of students in computer science and engineering who will later contribute to the development of computer systems in industry and research. No current scheme specialises software engineering best practice to the particular purposes of dialogue engineering, that is, to the development and evaluation of spoken language dialogue systems (SLDSs). DISC aims to develop a first detailed and integrated set of development and evaluation methods and procedures (guidelines, checklists, heuristics) for dialogue engineering best practice as well as a range of support concepts and software tools. The goals of dialogue engineering include optimisation of the user-friendliness (or cognitive engineering) of SLDSs which will ultimately determine their rank among emerging input/output technologies. DISC will contribute towards establishing dialogue engineering as a sub-discipline of software engineering, able to draw upon the rich variety of existing software engineering tools, methods and resources. DISC draws together some of the main actors in the national and European advanced SLDSs development projects that have been executed during the last decade.

Despite unquestionable progress, SLDSs development and evaluation is replete with unknowns and steps that are undersupported in terms of procedures, concepts, theory, methods and software tools. This situation continues to generate uncertainty about the potential of SLDSs technologies, their proper domains of application, their usability, the cost of producing them, their development time and the quality of products in both absolute and comparative terms. The DISC Action aims to take dialogue engineering and its successful commercialisation a significant step forward by pursuing the following specific objectives:

To carry out a detailed review and investigation of existing practices for a wide range of SLDSs and components development and evaluation.
To define a detailed current best practice scheme of methods and procedures for SLDSs and components development and evaluation.
To develop to the stage of industrial applicability a range of concepts, methods and software tools based on ideas and preparatory work at the partner sites.
To test methods, procedures, concepts and software tools on industrial and applied academic development projects to the extent feasible within the duration of the Action.

The industrial relevance of the expected results are:

Progress towards the integration of SLDSs best practice into software engineering.
Improved feasibility assurance of development projects (risk minimisation) and more exact feasibility assessment.
Improved procedures, methods, concepts and software tools.
Reduced development costs, accelerated development, improved maintenance and reusability.
Improved product quality and increased flexibility and adaptability.
Progress towards the establishment of dialogue engineering standards.
Improved guarantees to end-users that a product has been developed following best software and cognitive engineering practice. Enabling end-users to objectively assess different systems and components technologies against one another and choose the right product according to quality, price and purpose.

To achieve its objectives, DISC will (i) investigate a broad selection of state-of-the-art SLDSs to identify current practice and pinpoint its deficiencies; (ii) develop, test and disseminate a first reference methodology of best practice procedures and methods, and a toolbox of associated concepts and software tools.

Keywords: Best practice, interactive speech systems, spoken dialogue development and evaluation

Contact Information

Niels Ole Bernsen and Laila Dybkjær
The Maersk Mc-Kinney Moller Institute for Production Technology (MIP)
Odense University
Campusvej 55
5230 Odense M
Denmark
Phone +45 65 57 35 44
Fax +45 66 15 76 97
Email: nob@mip.ou.dk and laila@mip.ou.dk
URL: http://www.mip.ou.dk/nis/

Participants

The Maersk Mc-Kinney Moller Institute for Production Technology (MIP), Odense University, Denmark. (Contractor).
Centre National de la Recherche Scientifique (CNRS-LIMSI), Paris, France.
Institut für Maschinelle Sprachverarbeitung (IMS), Universität Stuttgart, Germany.
Department of Speech, Music and Hearing, Kungliga Tekniska Högskolan (KTH), Stockholm, Sweden.
Vocalis Ltd (VOCALIS), United Kingdom.
Daimler-Benz (DAIMLER-BENZ), Ulm, Germany.
Stichting Elsnet (ELSNET), Utrecht, The Netherlands.

CCS Natural Interactive Systems,
MIP, Odense University,
Campusvej 55, DK-5230 Odense M
Denmark

comments to: webmaster@mip.ou.dk
revised: October 28, 1997