HITO GUI

New GUI for HITO with better user experience. Available at https://hitontology.github.io/gui/.

Usage/Help/Summary

The GUI automatically shows every available path between the source and target node and then displays all instances which fulfill one path of the users' selection. Select the source node with the primary mouse button. Select the target node with the secondary mouse button.

Local Development

git clone git@github.com:hitontology/gui.git
cd gui
npm install
npm run dev

Open http://localhost:3000 in a browser.

Preprocessing

Perform the appropriate actions when the inputs change.

task	inputs	output	required action
download diagram	https://hitontology.eu/public/2024-05-hito_diagram.svg	img/2024-05-hito_diagram.svg	npm run download
transform diagram	img/2024-05-hito\_diagram.svg, scripts/preprocess.xsl	img/diagram.svg	npm run pre
generate hashes	HITO SPARQL endpoint https://hitontology.eu/sparql	js/pathHashes.js	open stats.html and copy into output

Contribute

Pull requests are encouraged. Please use Conventional Commit prefixes: fix, feat, build, chore, ci, docs, style, refactor, perf, test... For example: git commit -m "fix: useful commit message, resolves #123". Format before committing: npm run fmt or prettier -w ..

Background

The Health IT Ontology (HITO) is used to describe software products and application systems in health environments and studies describing them. HITO offers an RickView RDF browser, a SPARQL endpoint and faceted search for studies and software, but we weren't satisified with the search and exploration user experience. The master's thesis of Andreea Somesan presents mockups and wireframes for improving the user interface for the existing HITO faceted search, but we ultimately aim to explore a more expressive and intuitive search approaches to satisfy more complex information needs, such as which authors wrote studies concerning a given enterprise function. In contrast to keyword search and question answering, where users enter a query into a text field, our goal is to visualize the paths two given classes in HITO are connected and to allow faceted search on a path, showing the individuals along the way and letting users filter them. Earlier experiments with similar approaches in the SNIK project using shortest paths or automatically determining interesting paths did not yield satisfying results, so all non-cyclic paths are given to the user to select a single one. While the RDF graph is directed, the edge direction is sometimes arbitrary (e.g. "uses" vs "used by") so edges are traversed in both directions, treating the graph as undirected for the purpose of path finding.

Algorithm Idea

Pseudocode, automatically translated: Example: Which authors have dealt with a specific EnterpriseFunctionClassified?

• Click on Search Class for Search Input: e.g., EnterpriseFunctionClassified

• Click on Result Class for Results: e.g., FirstAuthor

• Automatically: All paths between Search Class and Result Class, where each class is touched at most once, are calculated and displayed in the meta-model image. Paths are referred to as "search paths." For example, the following search paths would be found:

  • EnterpriseFunctionClassified - EnterpriseFunctionCitation - SoftwareProduct - Feature Classified - FeatureCitation - Study - FirstAuthor
  • EnterpriseFunctionClassified - EnterpriseFunctionCitation - FeatureCitation - Study - FirstAuthor

• Click: Mark a class that must be included (or not) in the search path, e.g., FeatureCitation must be included.

• Automatically: Only the search paths that include/not include the chosen class remain, e.g.,

  • EnterpriseFunctionClassified - EnterpriseFunctionCitation - SoftwareProduct - Feature Classified - FeatureCitation - Study - FirstAuthor
  • EnterpriseFunctionClassified - EnterpriseFunctionCitation - FeatureCitation - Study - FirstAuthor

• Click: Mark a class that must be included (or not) in the search path, e.g., SoftwareProduct should not be included.

• Automatically: Only the search paths that include/not include the chosen class remain, e.g.,

  • EnterpriseFunctionClassified - EnterpriseFunctionCitation - FeatureCitation - Study - FirstAuthor
  • ...

• If only search paths remain that differ only by their edges (because there are multiple edges between two classes), then desired or undesired edges must be clicked or unclicked.

• If only one search path remains, then a table is generated with the names of the classes of the search path and the names of the connecting edge as column headers, e.g.,

  • EnterpriseFunctionClassified

    < functionCitation < EnterpriseFunctionCitation

    supports > FeatureCitation <evaluate feature < Study firstAuthor > FirstAuthor

• A search string can now be entered for the first column. In the first column, all instances of the corresponding class containing the search string are listed. In column n (n=2,...,LastColumn), all instances of the corresponding class are listed that are related to an entry from column n-1 via the specified edge.

Technologies Used

• JavaScript: Vanilla JS with ES6 modules
• Compatibility latest versions of Firefox and Chrome (earlier probably work a long while back but aren't targeted or explicitly tested)
• Testing: Vitest for unit testing
• Formatting: Prettier
• Linting Oxlint
• Dependencies: Directly imported from CDNs for easy deployment to GitHub Pages (no build step needed)
  • AG Grid (results table)
  • svg.js (SVG manipulation)
  • Cytoscape.js (graph functionality)
  • notyf (pop-up notifications)
• Build: No bundling process involved