Extracting Linguistic Speech Patterns of Japanese Fictional Characters using Subword Units

This study extracted and analyzed the linguistic speech patterns that characterize Japanese anime or game characters. Conventional morphological analyzers, such as MeCab, segment words with high performance, but they are unable to broken expressions utterance endings not listed in dictionary, which often appears lines of To overcome this challenge, we propose segmenting characters using subword units were proposed mainly for deep learning, extracting frequently occurring strings obtain their utterances. We weighted by TF/IDF according gender, age, each character show specific feature. Additionally, a classification experiment shows model outperformed conventional method.