An XML-based model-driven framework for geodata exchange

Abstract

Geospatial data (geodata in short) sharing is of great importance to promote and facilitate Geographic Information System (GIS) application. The adoption of numerous geodata formats and the existence of semantic discrepancies make geodata sharing a difficult task. Geodata translation, the transformation of geographic information from one data format to another and from one conceptual model to an intended conceptual model, is a key issue in geodata sharing. The current practice is to use black-box geodata translators to convert geodata from one format to another and, when required, one or more intermediate format(s) are used to reduce the number of translators to be developed. These translators are hard coded and cannot be customized to reconcile semantic differences between source data and the intended data representation. A number of geodata transfer standards have been developed, of which each was intended to serve as the intermediate format but they have not been as successful as expected. Moreover, to translate geodata via a standard format simply means using two black-box translators. No mechanism is provided in the process to address semantic discrepancy. The introduction of an intermediate format also makes data translation more subject to information loss. Recognizing these two barriers, this study aims to develop techniques for automating the generation of geodata translators and enabling high-level customization of geodata translation so as to facilitate semantic reconciliation. The promising model-driven methodology and XML technology are combined to dredge the long jammed channel for geodata flow. A model-driven framework for geodata exchange is proposed to empower the acquisition of the desired features. In the framework, a 4-layer conceptual model of data representation is adopted, which consists of instance, schema, meta-schema and meta-meta-schema layers in a bottom-up order. Each layer is assumed to have an associated formalism for representing data. Upper layer data specifies models for lower layer data. A proprietary data format is conceptualized as a specific 3-layer model of data representation and can thus be accommodated in the framework. In a layer, there is assumed to be means for specifying transformation from one model to another. Data translation is conceptualized as a top-down chain of model transformations starting from the meta-schema layer with upper layer transformation automatically guiding lower layer model transformation, thus model-driven. This procedure is entitled framework, because it can naturally accommodate various geodata formats without significant restrictions and is open to legacy, current and forthcoming systems. In addition to automating translator generation, it makes customization of translation possible and easy, both on a dataset-by-dataset basis and on a format-by-format basis. XML-based implementation of the proposed framework has been explored, prototyped and tested. XML (Extensible Markup Language) is adopted as the common syntax at instance layer. W3C XML Schema is adopted as the common and ultra formalism for specifying meta-schemas and schemas. On this basis, a means for declaratively specifying translation rules between meta-schemas (and schemas) has been devised and its associated processor developed. Meta-schema translation specifications can be processed automatically to generate enhanced schema translators, which can produce associated schema mappings in addition to translating source schemas to target schemas. Reconciliation of semantic discrepancy, if needed,can be performed by human operators through customizing schema mappings in a declarative way. Schema mappings (customized or not) can be processed automatically to generate data instance translators. Recommendations are given for further studies on completing the developed techniques and incorporating ontology technology in the proposed framework.