LE 2000 is a strategic study of the potential impact of language engineering on the European Community. Language engineering, in this study, includes all technologies, products and services which handle, process and communicate information through language. LE is intrinsic to most of the information and communication technologies (ICT) on which the social and economic life of Europeans is based. In LE 2000 we have distinguished language-based technologies from pure natural language processing. The latter is a narrow (if important) segment of language engineering. The scope of the LE 2000 is much broader. It aims to assess the potential impact of all types of technologies which handle language, both spoken and written.

The study was conducted in four phases, over a nine-month period. Phase I involved the identification and definition of categories to be used in the research; Phase II described technologies and products; Phase III assessed the potential impact of language engineering on the European economy; and Phase IV identified measures which should be taken, at both national and CEC levels, to achieve the potential benefits of language engineering.

The primary focus of the project has been on economic benefits. Overall, we believe a strategic programme to support language engineering in Europe could have a significant impact on the European economy. While some benefits could accrue to European suppliers of language-based products, the overwhelming proportion of potential benefit would be from improvements in the productivity and competitiveness of European enterprises who would use such technology. We estimate that, over a five year period, a full-scale effort to support the implementation of language engineering could generate more than 200,000 jobs within the Community, and could boost the output of the EC economy by more than 26 billion ECU.

Figure 1: Projected benefits of language engineering over a five-year period

These benefits would derive from the strategic exploitation of technologies which are, by and large, either currently available or very near to market. There is scope for support of R&D and pre-competitive research, as well as for programmes that will help define and expand markets which are ripe for exploitation. These include enhancements to basic telecommunications services, software markets, and pure language applications such as translation and language teaching.

The measures which would support the strategy outlined in LE 2000 are of a somewhat different nature from those which have been applied, historically, in support of language technologies. We believe there is scope for investment in technical research and the development of linguistic resources, and that there is an urgent need for more market-oriented programmes. In particular, we believe pilot implementations of new technologies and applications, rigorously controlled demonstrators with effective dissemination of results, and programmes aimed at meeting specific performance objectives, should all be priority approaches for support of language engineering. We also believe the CEC can play a critical (and absolutely necessary) role in promoting and enforcing standards within the markets affected by language engineering.

Overall findings

Our estimates of the potential economic benefits which LE could yield through the principal benefit scenarios identified in the project are presented below. The total benefit, over a five year period, could be as high as 210,000 new jobs within the European Community, and an additional 26 billion ECU (see Fig. 1) in output in member states. These benefits represent the upper limit on what we believe could be achieved by the most promising technologies in language engineering.

The impact on employment would, we believe, be favourable, although it would certainly imply a degree of restructuring within the white collar workforce. Because of the market stimulation and competitive advantage which would accompany improvements in efficiency, we believe the current level of employment could grow by 0.17% as a result of a vigorous campaign to maximise the benefits which language engineering could bring.

The impact on output would be considerably more robust, and would be greater, overall, than that on employment. The projected 26 billion ECU in additional output represents 0.55% of current levels of output. This growth would be, or course, in addition to currently projected growth rates which do not assume a substantial investment in language engineering.

Measures to promote the economic benefits

Achieving the benefit scenarios outlined in the study will, we believe, require measures at both member state and EC levels. They fall into six main categories:

• Standards: The Commission should, we believe, take an even more assertive role in the generation and support of standards related to the successful use of language technologies. Standards, both for technical interoperability and for benchmarking/performance, are urgently needed in localisation, translation, CALL, and document management systems of all types. In particular, we believe the Commission and member states should support an effort analogous to the Open Systems movement, which we have called an Open Language Standard. This Open Language Standard would set out requirements, which software suppliers would be encouraged to meet through procurement policies, and which would, for example, ensure that software was developed in a way which made it quick and easy to localise.
• Demonstrators and benchmarking programmes: We believe pilot projects which are carefully structured to demonstrate the commercial benefits of language engineering, are an absolute necessity both to stimulate the market and to lead market developments. Technology competitions, which may or may not be associated with pilot implementations, are a highly desirable method of pushing suppliers to meet the rigours of the market. The well known approach taken by DARPA in the US could be adapted to meet the needs of the European market, by designing projects which focus on specific language pairs, or on a set of languages as well as on specific technical benchmarks. We believe competitive R&D could eliminate many of the problems experienced in past EC-supported research, particularly of "pre- competitive" projects which prove to be non-starters in market terms.
• Lingware (linguistic processing and resources) development: There is still an enormous task facing the language engineering market if fully multilingual products and services are to be achieved, primarily in the full development of the linguistic resources ( such as multilingual terminology databases and dictionaries) which are necessary to support a wide range of applications. We believe creative intervention on the part of the Commission and member states will be necessary to supplement the considerable efforts which are already underway, and to correct biases in the market towards powerful owners of resources and dominant languages. Such resources will be critical to the successful development and implementation of localisation, translation, CALL, and document management applications.
• R&D subsidies: We believe there is still a role for public support of research and development in the technologies which will take language engineering forward in the future. Such efforts are particularly important to bring existing NLP technologies within commonly used IT configurations (such as 486 based PCs and commercial LANs) for translation, language teaching and document management. And while voice technology has had some support in Europe, we believe the market for voice technologies is hampered by a weak supply stream in Europe, which could be directly addressed by R&D support.
• Programme coordination: Many of the goals which are set out in our benefit scenarios can be promoted through Commission efforts to coordinate activities of other bodies, both public and commercial. Many of the language-specific recommendations set out in Section 5 should be executed at the level of the member states, but coordination of efforts to achieve consistency and parity is essential. In addition, the Commission has a unique role to play in supporting and coordinating the efforts of industry organisations to move forward in the use of language technologies.
• Support of small and medium-sized enterprises: We believe there is considerable scope for targeted support of smaller suppliers in the markets served by language technologies, and that in some areas coordinated multifaceted development of the supply stream is preferable to highly centralised, monolithic development programmes. This is particularly true for information services, where primary providers should be specialised and localised (even if secondary service suppliers are not). In addition, we believe that commercial translation and localisation can be effectively supported throughout the Community.

Figure 2 (unavilable) shows where each of these six categories of measures is important in helping to achieve the benefit scenarios of Figure 1. Note that each category is applicable across a number of benefit areas.