Outreach 2/5: build and manually review the contact list #43

Selection rules (proposed)

• Datasets: washdev, ploswater, uncnewsletter. datapapers excluded (data papers publish their data by definition).
• A paper qualifies when das_type contains "request" (case-insensitive) — catches the coded category on request plus the few rows carrying the verbatim statement.
• Contact email: correspondence_author_email, falling back to first_author_email (the fallback matters: it recovers 26 of the 44 uncnewsletter papers).
• Link: paper_url, falling back to https://doi.org/<doi>.
• One row per (email, paper); the send step groups rows by address so nobody gets more than one message.

Yields 173 papers → 170 contactable rows → 153 unique addresses, with 3 papers lacking any address (parked in a missing_email sheet for manual address hunting via ORCID/institutional pages).

Known data-quality problems the review must catch

• Scraped address typos: e.g. a.aprker@cranfield.ac.uk alongside the correct a.parker@cranfield.ac.uk — same person (Alison Parker), two "unique" addresses. Also seen: @ncsu.ede-style TLD typos. Regex validation can't catch these; eyes can.
• Greeting heuristic limits: "Dear Dr {last name token}" (with Jr./Sr./III suffixes stripped) mishandles compound surnames (e.g. Spanish double surnames greet by the second surname) and assumes a doctoral title. The greeting column must be editable and used verbatim by the send step.
• Same person, multiple addresses across datasets/papers — dedupe by person, not just by string-equal email.
• Stale addresses: some papers are 10+ years old; expect bounces.

Follow-up tasks for @larnsce

• Approve the selection rules above (or amend — e.g. exclude papers older than some year?).
• Review all 153 greetings and addresses in the generated CSV (an hour of focused work; this is the step that prevents embarrassing sends).
• Merge obvious same-person duplicates (the Parker case) and fix visible typos.
• Decide whether to hunt addresses for the 3 missing-email papers.

A draft implementation (list builder + generated CSV) exists on branch claude/automated-author-email-outreach-amipwb.

Refs #41