Frequencies and pointwise mutual information (PMI) of news topic categories by clusters for modern and old-fashioned racism.

The category names in the Topic Category columns show the characterization of each cluster (PMI > 0). Category names were sorted in descending order by PMI. PMI is defined as follows: PMIk(c) = log2 P(c|k) − log2 P(c), where c is a topic category and k is a grid cluster. PMIk(c) > 0 denotes the category c characteristic under the given condition cluster k. The frequency of topic category c is the ratio of the category in each cluster (P(c|k)). International news was the most frequent category among all clusters and both types of racism because we classified the news grids related to South Korea and North Korea as international news. This news was not about Zainichi Koreans (i.e., Korean residents in Japan), but the expression of modern racism, which is about Zainichi Koreans in specific, was observed.