A Sinhala chatbot for user inquiries regarding Degree Programs at University of Ruhuna

Abstract

Computer based chat system is very popular in the modern world. It can be either HumanHuman chat system or human-machine chat system. Human-Machine chat systems are wellknown as chatbots. There exists some popular and advanced chatbots which are capable of performing almost naturally. But most of the chatbots are only available for English Language. But the set of population who are incapable of getting the maximum use with the available chatbots due to this language barrier can miss the chance of getting support from conversational agents/chatbots. And also they don’t have to have a formal conversation when chatting with a bot, otherwise users may avoid chatting freely, due to the diffidence to ask questions. This research targets on coming up with a solution as a Sinhala Chatbot applicable for a specific domain. A generic chatbot has a high chance to be not responding for most of the questions, making it unusable and ineffective. Therefore, when having a domain specific chatbot, the users can directly ask the questions and solve their problems referring the exact point without wasting the time searching the internet. So, developing ‘A Sinhala chatbot for user inquiries regarding Degree Programs at University of Ruhuna’ mainly targets to solve two problems by being language specific (Sinhala) & domain specific. Rather than a chatbot answering simple questions, this chatbot targets to identify the intent of the user asking the question and respond accordingly. Therefore, this chatbot can handle contextual conversations. That way the reliability & the acceptability of the responses become high with respective to a traditional rule based, template-based or keyword-based bot. The chatbot is developed using RASA platform which is an open source dialogue management tool. RASA Core and the RASA NLU are the two main components of this framework, where RASA Core provides its users for more advanced dialogues / chats. It allows training the dataset in many ways such as interactive learning and supervised machine learning. RASA NLU takes care of natural language processing tasks such as intent classification and entity extraction in chatbots/dialogue models