Natural Language Processing Based Dialog System Generation and Management Platform

In this paper, we describe a dialog system generation platform that can help people without technical backgrounds to create chatbots in different domains such as banking, insurance, tourism etc. Using this system chatbots ranging from QA bots to conversational chatbots that can have more complex dialogues can be created. Using general Turkish datasets, a language model was trained with ELMo architecture to ensure domain independency. Intent detection models employed by chatbots created by this platform uses this language model for training. Besides, dialogue act classification is used by conversation policy manager to make conversations consistent. Lastly, platform includes an analytics tool that uses clustering algorithms to make analysis based on previous chatbot conversations.