Fig. 1: Number of topics (initial and final) in the iterative procedure of topic detection and clustering of the consensus matrix C.
From: Topic detection with recursive consensus clustering and semantic enrichment
Each line represent a different corpus, the larger is a corpus the higher usually is the number of topics. The black triangles illustrate the special values of the initial topics that oscillate without converging to a fixed point. (right) The schema of the initial, final topic plot: the number of topics growths till a given level then drops for lack of words to fill the individual topics. The highest level is the stability area of the topic partitioning. Each point of the curve in figure “1b” is the average of the points corresponding to the same value of the “Initial number of topics" in figure “1a''.
