#Future work
Crawl more onion domains
Find topics within already found clusters
Explore other possible interesting features -> weighting based on html tag
Improve topic model
Experiment with TSNE
Use word2vec word similarity model for country detection (so it also works on synonyms)
Viz:
(interactive) Concept graph per topic
Add border to slices in piechart that represent legal or illegal cluster
(interactive) Map per cluster (colored from low to high presence)
Foamtree for clusters within clusters