Detecting Signs of Depression for Using Chatbots – Extraction of the First Person from Japanese

Abstract

In a fast-paced urban life and an increasingly competitive environment, social pressures are increasing and people cannot help but put the pressure on themselves. So many people now suffer from depression. Depression is becoming a modern social dis-ease. If the bad mood can be resolved in time or medical attention can be sought, recovery is quick, but if it is left un-checked, it often leads to tragedies such as self-harm or suicide. The argument of this paper is not based on medical findings (CONTENTS) as in the past, but more on signs that can be detected from the content and style of users’ writing. For example, depressed people use too many negative mood words, especially negative adjectives and adverbs; depressed people use a lot of first-person singular pronouns. All these can be seen as the writing style of depressed people. It can be seen as a major characteristic of depression, in this case indicating less interest in others. The objective of this study is, therefore, to automatically extract the signs of depression from the chat texts, and to use these features to detect signs of depression from the chat texts. By detecting the signs of depressed patients in their daily conversations through a chatbot based on a diagnostic method for depression, we aim to enable people with signs of depression to be advised to see a doctor at an early stage.