ELSNET is creating a technology roadmap for the field of Human Language Technology (HLT) and publishing it on the Web, to serve as a major resource for the HLT community describing the state of the art and a vision of its future progress. This vision will continously be extended, refined and modified by researchers and companies active in HLT.

    Technology roadmaps are never finalized. Hence, this web page provides a state of information that demands correction, refinement, and update. The scientific and industrial community is invited to comment on the roadmap and suggest extensions and modifications through public contributions to the moderated Roadmap Forum. ELSNET will compile suggestions and create new versions of the roadmap on this web site.

    Results from various roadmap workshops under the auspices of ELSNET will be included.

The current state of this Roadmap extends work done for the ELSNET/Enabler Workshop that took place in Paris on Aug. 28-29, 2003, aiming at predicting the future neeed and development of linguistic resources. As a starting point, we included (in red colour) the applications identified at the DeepThought Roadmap workshop (see corresponding roadmap) and linked up to them relevant resources and their interpreters.

We distinguish three lanes, one for resources, one for processing components and one for applications. A resource is linguistic data that can be "taken" (re-used) by interpreters. A processing component is a re-usable program that can interpret resources and that forms part of an application system. An application is an HLT-based functionality that is of significant benefit to end users.

A Roadmap cannot be valuably designed by a few experts. It needs agreement by many, and this we want to achieve by soliciting your valuable suggestions for modification and extension.

The roadmap is organized into milestones and relations between them. In order to jointly work on a Roadmap, it is essential to agree about what is a milestone and what is a relation. Here is a proposal for how to define milestones and relations between them.

We are aware of the fact that the current version does not always follow these guidelines. In particular, we did not substantiate our claims in which years the resources and the technologies will be available. Rather we wish to leave this open to discussion. Only the applications depicted in red (taken over from the DeepThought Roadmap) are timed with good reason.

We look forward to the updcoming Roadmap Meeting during LREC 2004 in Lisbon on May 30, for which extensions and improvements of this roadmap are planned. We hope that the current state of the roadmap provides a good starting point for you to feel inspired to send us your comments and suggestions.

Nicoletta Calzolari, Steven Krauwer, Stephan Busemann