BERTopic's full visualization suite
Every chart BERTopic produces from the 100K fit. Not part of the talk narrative — kept here for anyone who wants to explore the topic model directly.
visualize_topics_over_time · a century of thematic evolution
Topic prevalence across decades. Superheroes explode post-2008, Westerns fade after 1975, space sci-fi reinvents itself every decade.
visualize_heatmap · topic similarity matrix
Cosine similarity between every pair of topic centroids. Dark stripes are tight thematic families; light regions are orphans.
visualize_hierarchy · dendrogram of topic merges
Agglomerative-clustering tree over topic embeddings. Shows which topics collapse together first as you reduce resolution.
visualize_barchart · top c-TF-IDF terms per topic
The "before LLMs" representation — grids of the most distinctive terms per topic. Useful for sanity-checking Claude's labels against the raw signal.
visualize_topics · all 253 topics as a scatter
Topics as 2D points sized by document count. Hover for the top-5 c-TF-IDF terms. Open full-screen (heavy embed).
5K topics over time
The smaller-scale version of the temporal chart. Useful as a side-by-side comparison with the 100K version.