Semantic similarity
Every published dataset placed so that similar ones sit near each other. Choose what “similar” means from the dropdowns: two dimensions, three methods each. The reasoning behind the selected method, including what it gets wrong, is printed below the plot. Click any point to see the dataset and its nearest neighbours under that method.
Similarity map
Click a point, or search above, to see the dataset and its nearest neighbours under the selected method.
How this method works
Written next to the implementation in
analysis/python/cafedash/similarity.py and carried into the data file,
so the text and the code cannot drift apart.
What the model found
Diagnostics
Parameters and fit statistics for the selected method. The silhouette width measures how well the clusters separate under this metric: it runs from −1 to 1, and values near 0 mean the collection is a continuum rather than a set of distinct groups. Text collections usually are, so a low value here is a description of the data and not a defect.