WELD Paradigm

Towards a WELD Charter: a vision for language documentation

A Charter for the WELD paradigm would include at least the following five benchmark principles of comprehensiveness, efficiency, state of the art, affordability and fairness:

Language documentation must be comprehensive. In principle this means that language documentation must apply to all languages. But economy is a component of efficiency, and priorities must be set which may be hard to justify in social or political terms: if a language is more similar to a well-documented language than another language is, then the priority must be with the second.

Language documentation must be efficient. Simple, workable, efficient and inexpensive enabling technologies must be developed, and new applications for existing technologies created, which will empower local academic communities to multiply the human resources available for the task. A model of this kind of development is provided by the Simputer ("Simple Computer") handheld Community Digital Assistant (CDA) enterprise of the "Bangalore Seven" in India (see http://www.simputer.org/), which could easily be incorporated into Eurpean and US project funding.

Language documentation must be state-of-the-art. In addition to using modern exchange formats and compatibility enhancing archiving technologies such as XML and schema languages, efficient language documentation requires the deployment of state of the art techniques from computational linguistics, human language technologies and artificial intelligence, for instance by the use of machine learning techniques for lexicon construction and grammar induction. The SIL organisation, for example, has a long history of application of advanced computational linguistic methodologies (see www.sil.org), and more research is needed here.

Language documentation must be affordable.In order to achieve a multiplier effect, and at the same time benefit education, research and development world-wide, local conditions must be taken into account. Traditional colonial policies of presenting "white elephants" to local communities which must be expensively cared for and then rapidly become dysfunctional, must be replaced by inexpensive dissemination methods - at third world Internet prices, it can cost hundreds of Euros to download a large, modern software package (not counting landline interruptions), and net-based registration and support is unthinkably costly, as is wireless data transfer.

Language documentation must be fair. If a language community shares its most valuable commodity, its language, with the rest of the world, then the human language engineering and computational linguistic communities must do likewise, and provide open source software (also to reap the other well-known potential benefits of open source software such as transparency and reliability). The Simputer Public Licence for hardware and the Gnu Public Licence for software are useful references. The development and deployment of proprietary software (and hardware for that matter) and closed websites in this topic domain is a form of exploitation which is ethically comparable to other forms of one-way exploitation in biology and geology, for example in medical ethnobotany and oil prospecting.[cont]