EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Lobsters·Aug 7·5 min read
social media rabbit holes, clusters, and the relative mixing times of random walks
TL;DR
A study of Twitter's tweet co-occurrence reveals distinct clusters of content, highlighting a tighter right-wing community compared to a more dispersed left-wing presence. This clustering impacts how users are recommended accounts, leading to more pronounced 'rabbit holes' for right-wing users.
✦ Why It Matters
Consider how content clustering affects your social media strategy and audience engagement.
Key Takeaways
- Analyze content clusters to understand audience behavior.
- Adjust recommendation algorithms to account for clustering effects.
- Explore user engagement strategies based on community dynamics.
- Monitor how clustering influences information spread on social media.
How It Works
The study utilized a co-occurrence matrix to quantify how often users tweet about the same domain. By applying PCA and t-SNE, the data was visualized in two dimensions, revealing distinct clusters of content.
The tightness of these clusters indicates how closely related the content is, affecting user recommendations.