Goal-Oriented Conversational System Using Transfer Learning and Attention Mechanism

Goal-Oriented (GO) conversational system or GO chatbot is making a significant difference in information availability and customer satisfaction in different business areas. This article describes a GO chatbot model, where transfer learning and attention mechanism are used. Usage of these deep learning techniques widens the application areas of the GO chatbot. Application of transfer learning allows the GO chatbot to transfer the common knowledge of one domain to another domain, which solves the problem of inadequate data for a particular domain. On the other hand, the attention mechanism helps the model to perform domain-specific chatting. The proposed model produces better results on previous research work datasets and also for a newly introduced organ transplant information dataset. Two main contributions of this research are, using transfer learning and attention mechanism for GO chatbot and introducing a new dataset for organ transplant information.