Content Discrimination of Comics Based on Users’ Reviews

The objective of our research is to develop a search system to support the exploration of comics. To achieve this objective, it is necessary to obtain content information from comics, such as the world setting and characters. Consequently, we propose a method that obtains content information about comics from their reviews on the Web. Specifically, the proposed method determines a set of keywords that reflect the content of target comics in an appropriate manner and visualizes the relationships between comics by connecting them using their common keywords. Term frequency-inverted document frequency (TF-IDF) and latent Dirichlet allocation (LDA) algorithms are used to determine these keywords. TF-IDF determines keywords that reflect explicit information in the reviews, while LDA determines keywords that reflect implicit underlying topics in the comics. The results of two evaluations conducted to confirm the performance of these algorithms demonstrate that TFIDF helps to obtain meaningful keyword sets, while LDA currently produces relatively unclear keyword sets.