The Declaratron, semantic specification for scientific computation using MathML

We introduce the Declaratron, a system which takes a declarative approach to specifying mathematically based scientific computation. This uses displayable mathematical notation (Content MathML) and is both executable and semantically well defined. We combine domain specific representations of physical science (e.g. CML, Chemical Markup Language), MathML formulae and computational specifications (DeXML) to create executable documents which include scientific data and mathematical formulae. These documents preserve the provenance of the data used, and build tight semantic links between components of mathematical formulae and domain objects—in effect grounding the mathematical semantics in the scientific domain. The Declaratron takes these specifications and i) carries out entity resolution and decoration to prepare for computation ii) uses a MathML execution engine to run calculations over the revised tree iii) outputs domain objects and the complete document to give both results and an encapsulated history of the computation. A short description of a case study is given to illustrate how the system can be used. Many scientific problems require frequent change of the mathematical functional form and the Declaratron provides this without requiring changes to code. Additionally, it supports reproducible science, machine indexing and semantic search of computations, makes implicit assumptions visible, and separates domain knowledge from computational techniques. We believe that the Declaratron could replace much conventional procedural code in science.