gsoc-2026
GSoC 2026 Week 2: Backend, Testing, and First Amharic Mappings
This week moved the Amharic DBpedia project from planning into implementation: backend infrastructure, dump tooling, browser tests, UI additions, and the first three concrete template-to-ontology mapping contributions.
2 May 30, 2026 - June 5, 2026
The main theme of Week 2 was building usable foundations. I worked on the backend services that will support future mapping automation, added early API endpoints for dump inspection, improved the site with test coverage and new UI sections, and also contributed real Amharic Wikipedia templates and DBpedia mappings. This combination was important because the project needs both engineering infrastructure and real ontology content to move forward.
Backend foundation
LangGraph-ready
Initialized the backend infrastructure that the mapping agent and dump workflows can build on.
API surface
Sanitize + preview
Added sanitizer and dump preview endpoints so users can inspect input before extraction.
Quality gate
Chromium + mobile
Added Playwright coverage for core website components across desktop and mobile layouts.
Ontology work
3 domains
Created Amharic templates, example wiki pages, and DBpedia mappings for three entity types.
Implementation work
Backend infrastructure for the mapping workflow
This week I started turning the project from a static UI into an application that can support real Amharic DBpedia workflows. The backend now has a foundation for LangGraph-based orchestration, which gives us a place to coordinate future steps such as dump loading, candidate extraction, ontology lookup, mapping validation, and human review. I also added sanitizer and dump preview endpoints, because before we can ask users or agents to process a dump, we need a safe way to inspect and clean what is coming in.
Statistics API backed by the extraction framework
I implemented a FastAPI endpoint that calculates statistics for a selected Amharic dump. Internally, this endpoint connects to the DBpedia extraction framework instead of inventing a separate parser. That matters because the website should reflect the same extraction behavior that the DBpedia ecosystem already uses. The endpoint gives the UI a practical way to answer questions like how much content exists in a dump and what kind of extraction output we can expect.
Website confidence through browser tests
I added Playwright tests for the website so that every important component is exercised in Chromium and in a mobile viewport. The mobile pass is important for this project because community contributors and readers may not always use a desktop machine. The goal is not only to catch broken routes, but also to protect the usability of the Amharic DBpedia UI as new sections are added.
New Amharic DBpedia UI sections
I added a news section to the new Amharic DBpedia interface so project updates can be surfaced directly inside the site. I also integrated a Tentris UI entry point, which is useful because Tentris is part of the RDF and knowledge graph exploration ecosystem. Together, these additions move the website closer to being a practical project hub rather than only a landing page.
Templates, wiki pages, and DBpedia mappings
The most concrete ontology progress this week was creating three Amharic template families and connecting them to real example pages and DBpedia mapping pages. I chose Scientist, Sportsmen, and Writer because they are understandable entity types and each one gives us a useful pattern for mapping person-related infobox fields into DBpedia.
Scientist
Created the Amharic scientist information template, added it to a real wiki page for Timnit Worku, and created the matching DBpedia mapping page.
Sportsmen
Created the Amharic athlete information template, tested it on the Abebe Bikila page, and mapped the template fields to DBpedia ontology properties.
Writer
Created the Amharic writer information template, added it to the Bealu Girma wiki page, and created the corresponding DBpedia mapping.
Why this week mattered
By the end of Week 2, the project had a stronger technical base and a clearer content path. The backend can begin to support agentic workflows, the UI has more practical project surfaces, the browser tests reduce regression risk, and the mapping work gives us real examples to validate against. That is the bridge I wanted for this week: connect infrastructure work with visible Amharic DBpedia contributions.