We found that it is difficult to reason about the linguitic characteristics of MCI person related to usage of emotional words and in this paper, we analyze the fluctuation of emotion during a talk by using an emotional vector which is extracted from the words in the story. We use pre-trained BERT model after fine-tuing by using WRIME language dataset.
First we find that the emotional intensity is decrease according to age and especially in case of MCI. We also find that the intensity of complex or social emotions can be kept even though MCI person.
Second, we trace the fluctiation of emotoin estimated from words in the story per 8 words. As a result, MCI person tends to fluctiate more than elder healthy person or young person.
