Renaissance Drama — Topics
A topic map of early modern English drama, built from ~500-word chunks of every play in the corpus: what each stretch of a play is about, which of those concerns recur across works, and how they are distributed across genres. Every number on the site leads back to the chunks it was computed from.
17,422chunks
15,370in the model (one edition per work); 2,052 placed from other editions
580editions
516works
63topics
53in the genre comparison (included)
53 / 0confirmed / candidate themes
1520–1641publication years
Every chunk as a point, coloured by topic; filter by genre, publication decade, author, title, play type, company or theater (filters combine); click a point for its page.
Every edition with its chunks in order, topic composition, DEEP metadata and other editions of the same work; searchable.
All clusters with keywords, works, typical chunks and full rosters — grouped by the researcher's decision: in the genre comparison, context only, or pending.
How the cross-work themes are distributed across comedy, tragedy, history, tragicomedy, moral, romance, pastoral and masque.
Within comedy, tragedy and history: how the topic shares of Shakespeare's works, and of every author with five or more works, compare with the rest of the genre — with a random same-size reference, and the works behind each difference.
The corpus by estimated first-performance / composition decade: sample size per decade, mean topic shares within comedy, tragedy and history, and a way to trace any cell back to its works.
Corpus, chunking, name masking, embedding model, representative editions, clustering and the choice of scheme, keywords, review, typicality and aggregation — with the parameters actually used.
What a topic page shows
The three keyword lists, the size of the cluster and how far it is concentrated in one work, its genre mix, the most typical chunk of each leading work, keyword-highlighted excerpts, and a sortable roster of every member chunk with its typicality.
What a chunk page shows
The play, author, publication year and Annals performance date, company and theater where DEEP records them, the topic with its typicality, then the full regularized text with the topic's distinguishing words marked and links to the previous and next chunk; the metadata table (edition, TCP and DEEP ids, node positions, topic in the other seeds) and the original spelling are folded below it.
How to read the numbers
Of the 63 clusters, 53 are included in the genre comparison, 6 remain undecided after review, and 4 are retained for context only. In the representative-edition sample, 9,153 of 15,370 chunks (59.6 %) are unassigned by HDBSCAN and remain in the denominator; the included topics account for 36.9 % of that sample's words. The other editions (2,052 chunks, 1,320 of them unassigned) remain available for browsing but do not contribute to these statistics. No chunk is ever deleted.