Question classifier in hierarchical architecture for chatbot

Natural language is a great user interface for human since human use it every day in conversation with others. However, it is very hard to create a computer program that understands natural language since it is ambiguous. This condition is true for developing NTU chatbot where chat bot needs to understand the question from users and reply quickly like a chatting conversation between human. Even though technology such as machine learning is increasingly growing, machine learning is still not good for multiple domain problems while there are several NTU related topics that need a specific approach for processing query. This is the reason why the concept of hierarchical approach is introduced to develop chatbot that distributes users’ question to specific components that process the query and reply to the respective users. One component of hierarchical approach is question classifier model that detect the topic of users’ question and passes the question to another component that handle the respective topic. The model is built from neural network model trained with possible users’ questions. In this thesis, the current implementation of question classification model is discussed, including its advantages and disadvantages.Bachelor of Engineering (Computer Science