Introduction to the Special Issue on Software Architecture for Language Engineering

Every building, and every computer program, has an architecture: structural and organisational principles that underpin its design and construction. The garden shed once built by one of the authors had an ad hoc architecture, extracted (somewhat painfully) from the imagination during a slow and non-deterministic process that, luckily, resulted in a structure which keeps the rain on the outside and the mower on the inside (at least for the time being). As well as being ad hoc (i.e. not informed by analysis of similar practice or relevant science or engineering) this architecture is implicit: no explicit design was made, and no records or documentation kept of the construction process. The pyramid in the courtyard of the Louvre, by contrast, was constructed in a process involving explicit design performed by qualified engineers with a wealth of theoretical and practical knowledge of the properties of materials, the relative merits and strengths of different construction techniques, et cetera. So it is with software: sometimes it is thrown together by enthusiastic amateurs; sometimes it is architected, built to last, and intended to be 'not something you finish, but something you start' (to paraphrase Brand (1994). A number of researchers argued in the early and middle 1990s that the field of computational infrastructure or architecture for human language computation merited an increase in attention. The reasoning was that the increasingly large-scale and technologically significant nature of language processing science was placing increasing burdens of an engineering nature on research and development workers seeking robust and practical methods (as was the increasingly collaborative nature of research in this field, which puts a large premium on software integration and interoperation). Over the intervening period a number of significant systems and practices have been developed in what we may call Software Architecture for Language Engineering (SALE). This special issue represented an opportunity for practitioners in this area to report their work in a coordinated setting, and to present a snapshot of the state-ofthe-art in infrastructural work, which may indicate where further development and further take-up of these systems can be of benefit