The purpose of this study is to visualize speech nuance using speech-balloon shapes in text-based communication. Though speech nuance disappears in a text communication, comic readers can recognize speech nuance from the speech-balloon shape, which expresses nuances, emotions, and intentions. Through the subjective evaluation experiments, we confirmed that the shapes of speech-balloon changed the emotion towards the phrases. In this paper, we focus on the relationships between the linguistic features of speech and speech-balloon shapes, and developed an estimation model for speech-balloon shapes with the linguistic features using Naive Bayes classifier. The effectiveness was confirmed especially in the estimation of “Explosion” and “Wave” shapes. The proposed method is expected to be used to visually represent a speech nuance that contains emotion in a text-based communication such as chat system and social networking service.