Development of a Chatbot Called Baymax to Support Suicidal People with Severe Depression, using Natural Language Processing and the DNN Neural Model

Depression is a mental illness that affects and is one of the main causes of disability. In Peru, it is an exponentially growing problem and in 2023, 280,917 adults, not counting children, were treated [3]. Therefore, technological tools such as chatbots are resources that support the care of people with depression. In this sense, a chatbot called Baymax was developed. This uses text processing by implementing a Deep Neural Network (DNN) architecture. The DNN was trained with 137 topics. Each topic contains between 5 and 10 questions with their respective answer. The DNN model demonstrated robust training performance, reaching a maximum accuracy of 0.8683. For this, different hyperparameters were tested in the model, reaching an optimal value with an architecture of 4 hidden layers and 2000 training epochs (this is described in Table III and Table IV). This model handles large volumes of unstructured data thanks to the propagation architecture of 4 layers in the hidden layer that allow learning with minimum loss. These characteristics allow learning complex patterns. This translates into a greater ability of the chatbot to understand the context and tone of questions and answers. The Baymax chat bot represents a significant advance towards creating intelligent tools that adapt to the individual needs of each user.