Fig. 8
Municipalities soft clustering on all the topics depicts the semi-supervised hard assignment obtained from a random membership matrix Z0 for m=3 groups using distance matrix D obtained from topic modelling with k=3 after 14 iterations. Left top Hard membership. Left bottom the conditional entropy of topics H(R|i) showing municipality-topic probability distribution uncertainty. Right top the CA bi-plot between Z∞ (illustrated by ∙) and the official classification (illustrated by \(\blacktriangle \)) with m=3. Right bottom The free-energy plot: decreases as the number of iterative steps increases (β=5,α=7)