Table 1 The thematic clustering algorithm

Given K initial clusters, the number n _U , and the set of prior probabilities {pr _d }_d∈D,

   1. Create a random partition ${\left\{V_i\right\}}_{i=1}^K$ of D with corresponding relations ${\left\{R_i\right\}}_{i=1}^K$.

   2. Compute p _t , q _t , and r _t for V _i .

   3. Compute α _t for V _i .

   4. For each cluster, select the n _U points for which α _t is the greatest to define the set U and the indicator values {u _t }_t∈T.

   5. Compute the probabilities {pz _d }_d∈Dfor each cluster V _i .

   6. For all d, assign a document to the cluster in which the document has the highest probability.

   7. Test for convergence. Terminate if converged.

   8. For a subset $D_s\subset D_{V_i}$, where the documents in D _s has the lowest 1% {pz _d } in V _i , re-assign to the clusters that have the second highest probabilities.

   9. Return to Step 2.