NZ political social media
This week, on Twitter, Reddit, YouTube and Facebook
Topics, framings, and platform patterns derived from the social
corpus — tweets, Reddit posts, YouTube videos, Facebook page
posts and their comment threads, plus Meta Ad Library disclosures — joined
on the same canonical topics as
discourse, press,
and audio. Cards rank by engagement-weighted volume
(likes + 2×shares + 3×replies), so a single viral
post lifts a topic the way it does on the platform itself.
Click any card for the topic detail.
Top topics — Last 7 days
The canonical topics with the most engagement on social media
in the trailing 7 days — a sliding window, so the cards keep working through the Monday handover.
The trend figure compares the selected period to its prior 4-week mean.
Stance bars (where visible) show how the AI classifier read the per-edge
framings; sentiment bars show whole-post emotional valence. Distinct
signals: a post can be supportive of a topic and negative in
tone (an angry supporter), and the lens disentangles them.
How this lens works
Posts are polled from each platform's public API on a cadence
(Twitter Basic tier hourly, Reddit every 30 min, YouTube and
Facebook daily). Each
post's text runs through a topic extractor that maps it onto
the same canonical topic taxonomy used by discourse + press +
audio — so a topic that spikes here will show up on the
other lenses too. Two AI classifiers run in sequence: a 5-class
stance classifier on each (post, topic) edge
(supportive / critical / dismissive / neutral-explainer /
mocking) and a 3-class sentiment classifier on each
post (positive / neutral / negative). The two together capture
texture short-form posts compress into very few words.
