Companion Care: An Intention-Based Assistance System Using Transformer

The rising prevalence of mental health issues has highlighted the urgent need for accessible and scalable solutions to support individuals in maintaining their mental well-being. In this project, we developed an intention-based assistance system that harnesses the power of a transformer-based model. By leveraging advancements in natural language processing (NLP) and machine learning (ML), our chatbot employs a textual and audio-based conversational interface to foster empathetic and non-judgmental interactions with users. This project deals with a generative model trained on a large dataset of conversational data and uses attention mechanisms to learn the relationships between words and their meanings. Our chatbot integrates Voice Emotion Recognition (VER), Text Emotion Recognition (TER), Intent Recognition, and a Language Model for user responses. This combination enables the chatbot to understand emotions in spoken and written language, accurately identify user intents, and generate contextually relevant replies. In voice emotion recognition, we leverage the efficiency of a 1D-CNN algorithm. For text emotion recognition, the robust BERT model is utilized. Intent recognition is achieved through the application of the SVM algorithm, ensuring accurate identification of user intentions. The generative model, responsible for dynamic responses, is powered by the Transformer algorithm. The effectiveness of the various models used is evaluated through multiple evaluation metrics. Overall, this project demonstrates the potential of using transformer models for building chatbots that can understand human language and provide intelligent responses.