Contact

Position:
Johns Hopkins University, Baltimore, MD, USA
Address
United States

Miscellaneous Information

Miscellaneous Information

Abstract Reference: 238
Identifier: I10.2
Presentation: Invited Speaker
Key Theme: 

Data Models for Astronomical Data Integration: Language, Patterns and Representations

Authors:
Lemson Gerard

I describe a model oriented approach to data integration in astronomy, developed in the context of the international virtual observatory effort. The main idea is that common data models should be used to describe the data holdings in distributed astronomical archives. This should allow users to infer information about these archives and possibly query those without needing to know the detailed structure of the individual archives themselves.
To enable this approach we were led to develop first a standardized modelling language for expressing the common data models, and second a mapping language that allows us to describe how instances of those models are stored in tabular data sets.
I will describe some features of the modelling language, which we named VO-DML and which shares the essential concepts of the class diagrams in UML. VO-DML is defined through an XML serialization format that is both machine readable and human writable with no great difficulty. A special feature of the language is that it allows reuse and interoperability of different models.
But the languages are only one piece of the puzzle; to support interoperability great care must be taken in the design of the models themselves. I will try to argue that the type of models that are best suited for this purpose are the ones corresponding to the domain models produced in the analysis phase of standard modelling approaches and I will show some analysis patterns that have been used in several scientific modelling efforts, ranging from cosmological simulations to genomics.