The project leader is Louis des Tombe at University of Utrecht. He wants a validation procedure and would like to cooperate with translation departments or service vendors. One of the goals is to measure translator productivity. For translation managers, the benefit of this is that benchmark procedures are created and norms are established. This will facilitate the selection of tools and work procedures to improve translator productivity.

If you want to collaborate, contact Louis at : destombe@let.ruu.nl.

EAGLES takes off in Europe

Started in February 1993, the EAGLES initiative will complement and expand efforts towards uniform data representation and will strive to improve harmonization of linguistic information. EAGLES is intended to accelerate the provision of standards for the development, exploitation and evaluation of large-scale language resources. It is expected to produce an initial set of wide-ranging proposals for European de- facto standards by mid-1995.

Imagine that you have just received a new 250,000 word Danish lexicon on CD- ROM. You load it into your multilingual natural language front end and hit Return. Surprise! It works. The lexicon conforms to ISO 38439, and interfacing is no problem.

Obviously,we are a long way yet from realizing this dream. Nevertheless, there are promising signs that developers of large-scale language resources--text and speech corpora, lexicons, grammars, and so on--are converging towards a common view on matters of format and representation. Nobody imagines that it will be possible to codify such convergencies into de jure standards for the foreseeable future. Current language technology is still changing too rapidly for that. But there are grounds for optimism in thinking that common guidelines for key language resources can be hammered out, to the extent that such re- sources will serve a variety of needs, rather than one specific application.

In order to further consolidate consensus-building amongst language researchers, EAGLES--the Expert Advisory Group on Language Engineering Standards--was recently launched within the framework of the CEC's DGXIII Linguistic Research and Engineering (LRE) Programme. The initiative aims to accelerate the provision of standards for developing, exploiting and evaluating large-scale language resources. The structure of EAGLES has resulted from recommendations made by leading industrial and academic centres, and by the CEC's Language Engineering strategy committees: More than 30 research centres, industrial organisations, professional associations and networks across the EC are donating labour towards the effort. With financial support from the CEC's LRE Programme, coordination is being carried out by the Consorzio Pisa Ricerche (Italy), while five Working Groups will execute the detailed work.

Background

The integration of natural language and speech processing into complex Information Technology applications has been hampered by the lack of generic technologies and of large-scale language resources. In Europe, there is particular concern as the Language Industries are mainly driven by small and medium-sized companies providing highly customized applications. These are slow to develop and modify, mainly because of the high costs of building the natural language or speech resources required for such applications. An associated problem is the diversity of formats and variable linguistic specificity of existing resources which hinder their reuse and indeed engender duplication of effort.

The aims of EAGLES are:

• to produce public, commonly agreed specifications and guidelines for specific areas of language engineering, based on pooling results from current European efforts and exploiting networks of expertise;
• to complement European R&D efforts;
• to promote the adoption of EAGLES results in future R&D ventures, and;
• to feed results to national and international standardization initiatives. Given the scope of the LRE programme, EAGLES will select only certain key issues of relevance to the standardisation process, and will initially be able to involve only a selection of relevant European and national research and development activities. EAGLES consists of five Working Groups, hosted by designated R&D centres; a Management Board; and a central support team. The definition of specifications and guidelines is carried out in the Working Groups. These are concerned with common methodologies for the following five areas: Text Corpora, Computational Lexicons, Grammar Formalisms, Evaluation and Assessment and; Spoken Language.