ajalil-123/rwanda-health-media-scanning-system

# rwanda-health-media-scanning-system -- Phase 1 (Text-Based News) A no-cost, semi-automated system that scans text-based health news relating to Rwanda and produces a reviewable shortlist for RBC's media review team. Built to run daily and weekly scans, each strictly scoped to that day's or that week's content only. This is **Phase 1**: local online news, international media, and research journals. Radio/TV/YouTube and social media (X) collectors are intentionally out of scope for this phase and can be added as new modules later without changing the existing structure (see `TECHNICAL_NOTES.md`). ## What it does 1. Collects raw items from three free sources: - **Google News RSS** -- broad discovery across the wider media landscape - **Direct RSS feeds** -- known priority outlets (starts with The New Times; add more as they're verified) - **PubMed E-utilities** -- research & journal articles mentioning Rwanda 2. Restricts results to a strict date window: the current day for a daily scan, the trailing 7 days for a weekly scan -- not an ever-growing pile. 3. Filters for health relevance using a maintained multilingual (English / Kinyarwanda / French) keyword list -- no paid AI involved. 4. Deduplicates near-identical coverage of the same story across outlets, robust to headlines being reworded/reordered between publishers. 5. Scores and ranks items for a proposed Highlights list (transparent heuristic: outlet count + keyword signal + recency). 6. Stores everything in a local SQLite database (`media_monitor.db`, a single file -- no server needed) and exports a **CSV** and **Markdown** shortlist for the editorial team to review, select from, and write summaries into, the same way they do today. ## What it deliberately does NOT do It does not write the report for you. Selecting which items matter and writing the summary text in the report's existing style stays a manual, editorial step -- this system's job is to replace the time spent visiting dozens of sites and ac …