Demonstration: Data Schema Diagrams
ViziQuer is a tool for visual exploration and queries over RDF datasets. It allows users to visualize the data schema diagrams and visually construct and execute rich SPARQL queries.
- ViziQuer website: https://viziquer.lumii.lv/
- ViziQuer Playground: https://viziquer.app/
- ViziQuer wiki: https://github.com/LUMII-Syslab/viziquer/wiki/
- Local setup options: https://github.com/LUMII-Syslab/viziquer-tools
Quick Links:
- Register the user at ViziQuer Playground.
- Create a project (
New ViziQuer Project) -
for the DBLP schema diagram Demo choose
DBLP Demoas the example project in the project dialogue
-
for a new project, based on any schema that is available in the server instance (ViziQuer Playground in the demo case), search for the schema name in the 'DSS Schema' field.
To work with a project available on internet or the file system, start a new project first (with any settings), then upload the project file.
- Click on any of the schema diagrams to explore them live in the tool environment.
- You can make changes to the diagrams. There are options for the data exploration from a class/node context in the diagram (work in progress)
- The option of creating new schema and query diagrams applies also to pre-built or imported projects (you are working with copies of the projects).
On the basis of the (abstract) data schema attached to the project, it is possible to create multiple schema diagrams of varying scope (the entire schema, or some its fragment) and granularity (modes of grouping classes, in-lining properties and showing (or not showing) non class-bound property sources and targets as nodes in the graph).
The schema creation panel is opened by 'Schema diagram' button in the project's main view 'Diagrams'.
The class and property sliders allow for simple selection of largest classes and properties (for classes a metric combining the individual count and incoming triple count is offered by default, this can be changed, e.g. to the individual count only).
Unchecking 'Simple top class selection' allows entering manual scope configuration mode, where the lists of included classes and properties can be adjusted.
To include property source and/or target nodes in the diagram, check 'Non-class nodes'. The list of sources for "orphan" properties is added to the 'Remaining classes' tab; these sources can be manually moved to 'Selected classes' to include in the diagram creation.
Principal steering means for diagram creation is 'Merging strength', set by default to 'Basic (5)', allowing merging into single node of some classes with slightly different incoming and outgoing property characteristics (thereby loosing some information about exact property ascription to classes). Other options are 'No' (no merging at all), 'Equivalent classes' (only classes with equal attribute and line characteristics), 'Weak (2)', Strong (10)' and Strong (20)' (can be used as an initial try for larger schemas).
The right-most column of parameters specify, when to show a property as a line in the diagram (opposed to in-lining both property ends), as well as whether to show the attribute and merged class lists in full.
The middle column deals with diagram presentation options (add all known super-classes to the classes in the selected fragment) as well as diagram presentation modes (add abstract super-classes, connect intersecting classes (may be interesting for finer class structure exploration) and use precomputed property ascription points in the diagram, or compute them anew from the schema structure data).
To specify creation of the initial DBLP schema diagram, keep 'Basic (5)' as the merging strength and on the right-most column set 'Lines with property no more than' to 4 (to force the visualization of rdf:type in in-lined form) and set 'Brief property lists' to 'No' (to avoid ellipsis ... in node class list and property list presentation); there is no need to touch the 'Extra parameters' tab.
'Compute nodes' (optional) gives a pre-view of the diagram contents in the form, where the numbers of the nodes and lines can be seen, as well as the descriptions of contents of the nodes (press the button repeatedly to see the new contents after you have made the changes to the parameters). 'Create diagram' creates the visual presentation of the diagram.
The Extra Parameters allow for some further steering of the schema diagram creation.
'Compact list of property target classes' allows introducing ellipsis ... when describing the "other end" nodes of an in-lined property end description.
'Grouping size factor' accompanies merging strength from 'Diagram parameters' tab; a larger factor means larger hesitancy in merging together larger classes, if opposed to smaller classes of the same relative similarity/difference.
'Duplicate non-class links in attributes' introduces extra attributes into nodes that are copies of diagram lines leading to non-class bound property targets. If the user determines that the node presentation of this property target is not worth it, the node can be safely deleted, without loosing the information about the property presence in the data set.
'Target merging' determines the strength of merging of property targets. 'Basic (sim > 0)' is the strongest merging options, where a transitive closure of the intersecting property ends is considered. 'Weak (sim > 1)' and 'Weak (sim > 2)' are weaker property target merging options.
After setting up the parameters (and checking the node preview, if desired), the schema diagram is created by pressing the button 'Create schema diagram'.
An auto-layout version of the schema diagram is created. For a better diagram visual quality, manual diagram re-positioning is usually helpful (this may range from small re-positioning of some visual out-liers to careful node positioning and line routing for publication-level quality).
The following diagram is obtained from the schema diagram configuration dashboard by keeping 'Simple top class selection' checked to consider all classes in the endpoint and checking 'Non-class nodes' to include nodes corresponding to the property sources and targets in the diagram.
On the 'Diagram parameters' tab we keep 'Basic (5)' as the merging strength and on the right-most column set 'Lines with property no more than' to 4 (to force the visualization of rdf:type in in-lined form) and set 'Brief property lists' to 'No' (to avoid ellipsis ... in node class list and property list presentation); there is no need to touch the 'Extra parameters' tab.
Initially, the diagram is auto-layout (try this in the live project), after what its positioning can be adapted (larger changes, including the element deletion, insertion and updating are possible, as well).
The node Target for sdo:url et. al. in the basic schema diagram has performed merging of the transitive closure of overlapping property target ends; thereby linking together also targets of properties that are not overlapping. The following weak target merging option provides a heuristics for reduced property end merging; still currently without strict guarantees that non-overlapping property ends shall not be brought together (the technical implementation of this feature can be added later for the sake of experimentation).
When splitting a transitive property end overlap closure, every involved property end (e.g., target) is attributed to one of the nodes; in all other nodes, where relevant, it is shown in the attribute notation.
The following diagram shows the schema presentation. based on classes only. It can be observed that the block of citations is not visually connected with the rest of the schema. The properties cito:hasCitedEntity and cito:hasCitedEntity are recorded in the schema for the cito:Citation class, as well as there is :omid property in the Publication class, still there is no clue that the targets of these properties may overlap; so establishing the links between the publications and their citations.
The facts of targets common for owl:equivalentProperty and datacite:usesIdentifierScheme, as well as :listedOnTocPage and :indexPage are also missing from the following diagram.
We have used a more detailed class presentation mode (Equivalent classes), if compared to the previous diagrams, to outline, what class-based information the previous diagrams were loosing (in essence, the potentially wider scope of :publisherAddress was admitted, as well as the superclass in the subclass relation was not identified explicitly (although it had emerged clearly due to bringing forward the name of the largest class in the union).
Furthermore, rdf:type relation has been shown visually; since its targets, if any, are always owl:Class, its in-lined presentation in the previous diagrams do not lose essential information.
We provide diagrams for other schemas. The diagrams can be reproduced live in ViziQuer Playground environment. If the project with the diagram positioning is not supplied, create a project with the respective (abstract) data schema, then follow the steps of creating a new schema diagram (starting from the 'Schema Diagram' button) in the new project environment.
The visual diagrams can be created also for other schemas available on ViziQuer Playground (more than 200 unique data set schemas); for some of the schemas the class-based visualization conveys the data structure well, while for others the node-based property end visualizations are essential.
In the case of large data schemas consider visualization of schema fragments.
The data schema (abstract structure) has been extracted from https://qlever.dev/api/olympics on May 09, 2026.
The visual environment is available on https://viziquer.app, search for "olympics" in the DSS Schema field in the New Viziquer project dialogue.
Don't forget to check the 'Non-class-nodes' box in the schema diagram dialogue, followed by unchecking 'Simple top class selection' and transferring all found property sources from 'Remaining classes' list to 'Selected classes' list.
The positioning of the diagram after the auto-layout is easy (this can be a good first exercise in mastering the diagram positioning skill).
The central source node keeps the diagram together. The data model designers are free not to assign class information to all resources.
In the following diagram of Aalto University institutional data (the schema extracted from https://data.aalto.fi/sparql on May 09, 2026) the property end presentation in a node form helps understanding the connection between bibo:Article and :Person .
Core part of the https://ideal-rdf.dbcls.jp/sparql endpoint (schema extracted on 09.05.2026) contains just a few classes. Still, the presentation of the schema in form of a diagram allows seeing the property connection patterns that are present in the data.
Endpoint https://sparql.api.identifiers.org/sparql (schema extraction 09.05.2026)
The schema does not rely on the property end visualization, yet it has been very helpful in communicating the endpoint structure to the domain experts.
