Research Projects

Background: Anxiety disorders are widespread mental health conditions that have a significant impact on a large number of individuals globally. Mental disorders, such as anxiety, rank among the top 10 leading causes of global health loss. In 2019 alone, an estimated 970 million individuals worldwide grappled with mental health conditions. An additional 76 million faced anxiety disorder, marking a 28% increase attributable to the COVID-19 pandemic. The prevalence of this disorder is significant, exerting a considerable influence on individuals' daily functioning, relationships, and overall quality of life. Conventional mental health services frequently encounter constraints, including long waiting periods, a scarcity of trained practitioners, and individuals' hesitancy to seek assistance due to concerns about being judged or misunderstood. Artificial intelligence (AI)-driven virtual assistant using deep learning techniques, and natural language processing, for diagnosis and classifying anxiety disorder represents a significant opportunity to bridge mental health care delivery gaps. Method: The development of the multimodal virtual assistant for anxiety disorder involves a comprehensive approach combining psychological and psycholinguistic literature insights with state-of-the-art AI technology. A thorough study of existing literature on anxiety disorder, and AI-powered mental health interventions will be undertaken. The initial steps include diagnosing the anxiety from the user conversation through text or speech and then further sub-classifying it. It entails the identification of evidence-based therapeutic approaches, such as mindfulness-based interventions cognitive-behavioral therapy (CBT), and other psychological therapeutic techniques. Subsequently, a dataset will be meticulously compiled in Hindi, integrating psycholinguistic, psychological knowledge, and crucial insights from the literature review. This dataset contains various parameters derived from the existing literature, such as the specific types of anxiety disorders, emotions, cause, essential strategies, and other relevant information for individual conversations. Afterward, an expert deep learning model will be created, utilizing the capabilities of a large language model to improve the virtual assistant's comprehension, empathy, user satisfaction, and interaction. This model will be been trained on a compiled dataset to ensure its efficacy in delivering meaningful interactions and assistance to users. The assessment of the virtual assistant's performance will be carried out using both automated and human evaluation metrics. In addition, human evaluation will be performed to assess the assistant's fluency, ability to identify information, capacity to provide comfort, and effectiveness in making suggestions during interactions. Later, a user-friendly and intuitive interface will designed for the virtual assistant, ensuring easy access through web or mobile platforms.