Mapping Physical Formats to Logical Models to Extract Data and Metadata: The Defuddle Parsing Engine

摘要

Scientists, fueled by the desire for systems-level understanding of phenomena, increasingly need to share their results across multiple disciplines. Accomplishing this requires data to be annotated, contextualized, and readily searchable and translated into other formats. While these requirements can be addressed by custom programming or obviated by community standardization, neither approach has solved the problem. In this paper, we describe a complementary approach--a general capability for articulating the format of arbitrary textual and binary data using a logical data model, expressed in XML-Schema, which can be used to provide annotation and context, extract metadata, and enable translation. This work is based on the draft specification for the Data Format Description Language and our open source Defuddle parser. We present an overview of the specification, detail the design of Defuddle, and discuss the benefits and challenges of this general approach to enabling discovery and sharing of diverse data sets.