Design and Implementation of Therapist Chatbot Using Encoder-Decoder LSTM

This paper presents the development and deployment of a therapy chatbot powered by an encoder-decoder LSTM model, a generative artificial intelligence approach for providing mental health support. The chatbot is implemented using an encoder-decoder LSTM model that is trained on a dataset containing some standard question answer pairs. The methodology involves data preprocessing, tokenization, vocabulary building, data vectorization, model architecture, training, and response generation. The encoder-decoder model is proficient at understanding the context of a conversation and creating responses that align with previous interactions. Additionally, the chatbot demonstrates the ability to generate grammatically accurate responses that also flow seamlessly. The generative-based encoder-decoder model also gives more promising results than retrieval-based and rule-based therapeutic chatbot models in context to training and validation accuracy.