Uniform Language Resource Access and Distribution

The reuseability and accessibility of lexical and linguistic resources often requires substantial programming overhead and detailed knowledge of the structure and nature of resource-specific information. Each resource has its own representation syntax and covers a particular subset of linguistic phenomena. This paper will discuss ways to overcome these barriers to resource reuse and presents a new method for distributing and accessing language resources within the GATE environment (General Architecture for Text Engineering). In this environment linguistic resources are modelled in an object-oriented way, yielding an initial set of resource-specific inheritance hierarchies. The exploitation of commonalities between resources leads to an incremental integration of their hierarchies into one common object model (whilst preserving the originals). The representation of objects within this model will be geared towards generally accepted linguistic concepts, which will be realised in a standardised format. This approach has several advantages: the linguistic knowledge is modelled in a conceptual and maximally uniform way; the linguistic resources are represented by maximally uniform data structures; overlaps and differences between resources can be identified by means of the common object model; the model provides a level playing field for the evaluation of resources. In describing parts of the model we will stick closely to the terminology and structures of the existing resources. This is not a new standards initiative, but a way to build more effectively on previous initiatives. The distribution mechanism for the object model will use WWW protocols and enable remote usage of resources without the necessity of