The study
Hierarchical clustering and DBSCAN on the same books, compared on two kinds of features: word weights, and meaning from a multilingual encoder chosen by a test of five. Each part opens with the finding in plain words; the numbers are underneath.
- Results at a glance
- The sample
- Preprocessing
- Distance
- Hierarchical
- Density-based
- Compared
- Stability
- What drives the groups
- The app’s model
- Accuracy
- Limits
Results at a glance
What the study set out to show, and whether it did. Each result sits beside the simple guess it has to beat; the targets were fixed before any result was seen, and a missed target is shown as missed. Where an interval is recorded, the thin line across a bar is the range the result would likely fall in on another sample of books.
Loading the results.
Loading the study results.