Project:Analytics/PubPeer

API

https://dashboards.pubpeer.com/docs/api#/operations/partner

Relevant parameters:

• page: start with 1 then iterate based on whether there are more results
• per_page: set at maximum value 300
• sort:
• published_at: concerns when the document was published; I only care about comments

Resources

• Wikimedia Cloud Services
  • Toolforge: project "pubpeer"
    • Don't think it's being used for anything
  • Cloud VPS: project "wikicite"
    • VM wikicite-refsdb-proc-1.wikicite.eqiad1.wikimedia.cloud
    • Trove DB: ouqdvgrbzf3.svc.trove.eqiad1.wikimedia.cloud

Process

Data collection and indexing

PubPeer Data (index_pubpeer.py)

• Initial Seed:
  • Starts from 2000-01-01
  • Uses an initial large date window (2000-01-01 to 2014-12-31) followed by 30-day increments
  • API query handles pagination (per_page: 300) and automatically reduces the time window if the result set hits the 10,000-record limit to ensure no data is missed
• Subsequent Builds:
  • Triggered via python index_pubpeer.py --update
  • Identifies the latest_comment_date from the local database and starts fetching from that date to the present
• Database Updates:
  • Updates the pubpeer_articles table
  • Fields: id_pubpeer (URL), id_doi, id_pubmed, id_arxiv, title (truncated to 250 chars), and time_last_comment

Wikipedia Citations (index_citations.py)

• Process:
  • Pulls current external links from Wikimedia Cloud DB Replicas for DOI (org.doi.), PubMed (gov.nih.nlm.ncbi.pubmed.), and arXiv (org.arxiv.)
  • Restricted to Main (0) and Draft (118) namespaces
  • Matches these links against the local pubpeer_articles table
• Database Updates:
  • wikipedia table: Stores language_code, mw_page_id, mw_page_title, and mw_talk_page_id. Page titles are refreshed during each run
  • citations table: Maps id_pubpeer to id_wiki_page
  • Stale Data: Automatically removes citations from the local database if the link has been removed from Wikipedia

Reporting and wiki updates

• Wiki Maintenance:
  • Automatically handles page moves and deletions via sync_wikipedia_titles.py before updating reports
• Report Generation (report.py):
  • Alerts Report: Lists new citations (time_last_updated_table IS NULL) or existing citations with new comments (time_last_comment > time_last_updated_table)
  • Most Affected Report: Lists Wikipedia articles with the highest number of unique PubPeer-commented citations
  • Article List Reports: Large table and alphabetical subpages (/By article/A, etc.) listing all matched citations
  • Frequency Report: Aggregates by PubPeer article to show which research is most cited across Wikipedia
• User Interactions (Dismissals): (pending working implementation)
  • The bot reads the current wiki report and compares it to its previous version
  • If an editor removes a row from a wiki table, the bot marks that citation as dismissed = TRUE in the database and stops reporting it

Database Schema (schema.sql)

• wikipedia: id, language_code, mw_page_id, mw_page_title, mw_talk_page_id, librarybase_id
• pubpeer_articles: id_pubpeer, id_doi, id_pubmed, id_arxiv, title, time_last_comment
• citations:
  • Links articles to wiki pages
  • cited_id_type: 1 (DOI), 2 (PubMed), 3 (arXiv)
    • 0 is reserved for "other"
  • time_last_updated_table: Tracks when the wiki report last included this citation
  • time_last_talk_page_post: (Reserved for future talk page notifications)
  • dismissed: Boolean flag for editor-driven dismissals

Post-Implementation Status

• Talk Page Notifications: Code includes fields for tracking (time_last_talk_page_post), but the active workflow currently focuses on centralized reports (Wikipedia:PubPeer/*) rather than automated talk page posting
• Frequency of Runs: Designed to be run periodically (e.g., via cron) using the --update flag for index_pubpeer.py
• Dismissals Not Recognized: The logic to recognize user removals of report entries does not work yet and has been disabled.