Tutorial: Monitor citation churn on a medical/scientific page

Goal

Track citation additions and removals on a high-traffic medical Wikipedia page to detect sourcing changes over time.

Steps

1. Analyze a medical page

refract analyze "COVID-19" --depth detailed -c

2. Export citation data as CSV

refract export "COVID-19" --format csv

This produces a CSV with citation URLs, add/remove timestamps, and revision IDs — ready for spreadsheet analysis.

3. Pipe to file

refract export "COVID-19" --format csv > churn-report.csv

Use case: identifying biased sourcing

Citation churn patterns reveal source instability. A citation added in one edit and removed in the next signals contested sourcing. Citation replacement (journal swapped for a news article) indicates editors updating their evidence base. Churn concentrated on one page section may point to a content dispute.

Example output

{
  "eventId": "d1e3f5a7b9c2048a",
  "eventType": "citation_added",
  "fromRevisionId": 1280090010,
  "toRevisionId": 1280090100,
  "section": "Vaccine efficacy",
  "before": "",
  "after": "<ref>{{cite journal |last1=Smith |title=...}}</ref>",
  "timestamp": "2024-11-20T10:00:00Z",
  "layer": "observed",
  "deterministicFacts": [
    {
      "fact": "Citation added in section Vaccine efficacy",
      "provenance": {
        "analyzer": "citation-tracker",
        "version": "0.3.1",
        "inputHashes": []
      }
    }
  ]
}

{
  "eventId": "2b4d6f8a0c1e3059",
  "eventType": "citation_removed",
  "fromRevisionId": 1280090100,
  "toRevisionId": 1280090200,
  "section": "Vaccine efficacy",
  "before": "<ref>{{cite journal |last1=Smith |title=...}}</ref>",
  "after": "",
  "timestamp": "2024-11-21T14:30:00Z",
  "layer": "observed",
  "deterministicFacts": [
    {
      "fact": "Citation removed in section Vaccine efficacy",
      "provenance": {
        "analyzer": "citation-tracker",
        "version": "0.3.1",
        "inputHashes": []
      }
    }
  ]
}

Notes