ABSTRACT
The Indian Psychiatric Society estimates that the number of mentally ill people in India has risen by 20% during this pandemic period. Lots of people seeking mental support end up in depression and many are still suffering. To provide Artificial Intelligence support, the authors proposed a multilingual chatbot enabling us to address this issue with the help of an interactive interface to the end user by psychological behavior pattern analysis. The proposed Co-Bot has well-designed architecture along with a structured knowledge engine that supports providing a language-independent interface with extendible scope to add further languages. The design consists of three components with a well balancing of cohesion and coupling. These components are a natural language processing (NLP) engine, a knowledge engine and an interactive interface enriched by a virtual avatar. It would be supported by a structured resource repository of coronavirus disease 2019 (COVID-19)-related information by means of medical knowledge and conversation dialog as well.
For language engines, a sequence-to-sequence (Seq2Seq) transformer model acts as our primary conversational model. This model forms the core of the NLP engine design and processes user requests and responds accordingly. A sentiment analysis model operates in the background. This model detects negative or depressive elements in the user's dialogs after conversation. Based on the accuracy of prediction, the patient would be recommended for the Patient Health Questionnaire-9 (PHQ-9) depression test to detect the degree of his/her depression to classify its intensity into one of the five categories, from mild depression to severe depression.
The second module of our system calculates the probability of a user contracting COVID-19 using parameters, including age, gender and symptoms such as fever and fatigue. The information is obtained from the user in the form of a questionnaire and thus calculates the probability as mentioned earlier. Based on the probability calculated from the system, it warns the user about having/not-having contracted COVID-19 and advises accordingly. The next module of our system is an interactive interface with a virtual avatar in a rich environment supported by virtual reality technologies with voice and language support.
