The function of the road map is not to impose anything on anyone, but rather to provide a broadly supported definition of a context in which to position the MT community's efforts, which would allow us to identify common priorities for joint activities in e.g. research, resources and training.

For this workshop we have invited papers that

• give critical analyses of the present state of the art in machine translation of written and spoken language,
• present visions of the future of machine translation, both from a technological and from an application point of view, or
• identify major milestones and challenges on our way towards the future, and/or ways to measure our progress along the road.

As the workshop was planned as a half-day event, we could only accommodate a limited number of papers. The nine papers that have been selected for this workshop cover broadly three different aspects of MT. The first group of papers (by Gerber, Schütz, Boitet, Farwell and Helmreich, Och and Ney) all address the question: where to go? Each of them sketches its own perspective of a potentially successful future for MT. The second group of papers (Macklovitch and Valderrábanos, Flournoy and Callison-Burch) deal with the role of interaction in MT, and make it clear that the topic may be old, but far from obsolete. The last two papers are different in that they do not specifically address the future development of MT, but rather remind us of two issues that we should not ignore when discussing the Big Problems and the Grand Challenges that lie ahead of us. The paper by Tsou and Kwong clearly illustrates that the notion of problem is relative: what counts as a non-problem when translating between Western European languages (translation of personal names) turns out to be a really hard problem when translating between e.g. Chinese and English. Probst et al draw our attention to the existence of low-density languages, where the lack of commercial interest may make it hard for speakers of those languages to get access to translation facilities and hence to the multilingual information society.

I hope that the confrontations of the various ideas expressed in the papers will lead to a fruitful workshop with lots of discussion, and to a common view on the future of MT. I would like to thank all contributors and the members of the Programme Committee for their efforts.

Steven Krauwer (ELSNET / Utrecht University, Netherlands) Chair

Niels Ole Bernsen (NIS, Denmark)

Paul Heisterkamp (Daimler-Chrysler, Germany)

Jose Pardo (UPM, Spain)

Pierre Isabelle (XEROX, France)

Arjan van Hessen (IP Globalnet / Twente University, Netherlands)

Antonio Zampolli (U. Pisa, Italy)

Hans Uszkoreit (DFKI, Germany)

Oliviero Stock (IRST, Italy)

Susan Armstrong (ISSCO / ETI, Switzerland)

Herman Caeyers (LANT, Belgium)

Nuria Bel (gilcUB, Spain)

Ed Hovy (ISI, USA)

ELSNET is the European Network of Excellence in Human Language Technologies,
hosted by the Utrecht institute of Linguistics OTS (UiL OTS), and funded by the European Commission's Human Language Technologies Programme.

Address: ELSNET

Trans 10

2512 JK Utrecht

The Netherlands

Phone: +31 30 253 6050

Fax: +31 30 253 6000