Schema analysis and shape extraction

[jmillanacosta]

Identify which schema extractor to use at the start of the workflow.

Identified strengths and limitations:

	VoID generator	RDF-config	sheXer
Overview of the tool	Extracts statistics from an RDF endpoint or a file	Automates SPARQL and schema diagram generation	Extracts ShEx and SHACL structure from an RDF graph
Documentation	Incomplete	Present	Present
Minimal requirement	Requires an RDF file or SPARQL endpoint; Graph must have triples with rdf:type predicates	Requires manually curated model.yaml and prefix.yaml files	Requires a Turtle file or a SPARQL endpoint
Data model representation	Object class-based file detailing all classes and properties	A set of tree structures representing classes and properties in YAML which can be exported as SVG	A SHACL or ShEx schema, optionally a UML diagram in PNG representing the shape graph
Process	Automated	Semi-automated	Automated, and optionally one can change the threshold used to accept patterns in the graph as shapes and I/O settings
Interpretability	Difficult; requires programming knowledge to understand organization of classes and properties	Easy; human-readable terms and graphical representation make the graph structure easily understandable	Easy; graphical representation facilitates quick interpretation
Readability	Human-readable terms (mapping ontology to labels); more compatible with programming language formats	Human-readable terms assigned for classes and properties	The diagrams and graphs are rendered with their URIs and no labels
Error reporting	Through Git issues	Through Git issues	Through Git issues
Implementation	Java (Native binary)	Ruby	Python
Limitation	Quadratic runtime for generating files (e.g., IDSM, OrthoDB); Not applicable for shape subclasses (e.g., In Rhea, where compounds are sub-classified into products, reactants, etc.)	Requires manual curation of input (Potential solution: integrate with VoID generators for semi-automation, for which a prototype exists)	The current algorithm is unable to detect specific patterns within the data; however, the developers are aware of this limitation.