Identifikasi Cuitan Berbahasa Indonesia yang Mengandung Unsur Depresi Menggunakan Metode Bidirectional Gated Recurrent Unit

Abstract

Twitter can be used to express and describe the emotions felt by the users. The emotions expressed through tweets on the users’ Twitter accounts can be associated with the mental disorders they are experiencing. One of the most common mental disorders is depression. The high rate of depression in Indonesia should be watched out for because depression can make a person unable to do their daily activities smoothly. Twitter can be a tool for early detection of depression through the tweets written by the users on Twitter. However, identifying tweets that contain depressive elements will require a long time and thoroughness if it is done manually. So, it is required to have a model that is able to automatically recognize a person's potential to experience depression and enable the person to have a proper diagnosis and treatment for earlier treatment. In this study, Bidirectional Gated Recurrent Unit method with word embedding FastText was used to identify tweets in Indonesian that contained depressive elements and did not contain depressive elements. The dataset used in this study was 7.000 data consisting of 4.480 training data, 1.120 validation data, and 1.400 testing data sourced from Twitter. This study produced an accuracy of 87%. Based on the results, it can be concluded that the system created using the Bidirectional Gated Recurrent Unit method and word embedding FastText is good at identifying Indonesian-language tweets containing depression