Visualization of group discussion using correspondence analysis and LDA in Ideathon
Makoto Sakiyama, Nobutada Fujii, Daisuke Kokuryo, Toshiya Kaihara · Procedia CIRP · 2020
Ideathon has attracted attention in recent years but the results of ideathons are not stable; good ideas are not often generated. Therefore, it is necessary to analyze and support the idea creation process in the ideathon. This research focuses on group discussions and analyzes the idea creation process by visualizing the discussions. Visualization is performed by combining correspondence analysis with LDA (Latent Dirichlet Allocation). From the visualization results, divergence of the discussion is evaluated quantitatively using cosine similarity. It is also revealed that group discussion with better evaluation includes higher divergence.