Vaccine policy and social listening in Taiwan: a dashboard

This post shares my independent project, a public, bilingual dashboard that tracks vaccine policy and public conversation in Taiwan every day. It covers:

  • What the dashboard shows for flu, RSV, Tdap and COVID-19 vaccines
  • How an AI pipeline collects, tags and summarises the signals

Link to the dashboard and the GitHub repo.

Why a radar?

Every vaccine season, the same questions come up.

What did Taiwan CDC announce this week? Are people talking about publicly funded or self-pay doses? Is a rumour or misinformation gaining ground, and where? The answers are scattered across press releases, media news sites, and various social platforms like Threads.

I wanted one page that answers these questions on a daily basis - in support of my marketing and communication strategy.

What’s on the dashboard

The dashboard has four tabs, all available in English and Traditional Mandarin:

  1. Today: an AI-written (Claude Haiku model) daily brief of the 4–6 most important developments in the last 72 hours. Each item comes with a confidence level, a “why it matters” line and links to its sources. Next to it are a watch list of upcoming dates, headline figures, and conversation volume with official actions marked on the timeline.

Taiwan Immunization Radar

  1. Policy: a timeline of Taiwan CDC and TFDA actions, such as press releases, letters to physicians and rumour rebuttals.

Policy

  1. Social listening: mentions by channel, a county-level map, spike alerts, a misinformation watch and a narrative explorer.

Social

  1. Methodology: sources and their live status, definitions, limitations and AI-use rules.

The filters combine freely: date range, vaccine, who pays, COVID-19 brand, and any of Taiwan’s 22 cities and counties.

How it works

Every day GitHub Actions runs a pipeline:

Step What happens
Collect Taiwan CDC, TFDA, Google News, PTT, Threads, Facebook page comments, Wikipedia pageviews using Apify
Tag Claude Haiku labels relevance, vaccine, who pays, place, issue and stance
Scrub Verbatim social text is deleted; only paraphrases, tags and links stay
Analyse Claude Haiku writes the brief, narratives and misinformation notes, then translates them
Publish Data is committed to git, so the history doubles as an archive, and the site deploys to GitHub Pages
This is an independent portfolio project, not affiliated with or endorsed by any company or agency named. Views expressed here are my own. This post was collaborated with Claude.