Authors: Chuthamanee Tangkathach, Vishnu Kotrajaras
Chatbots have been used to provide a different way of services on many websites since it can help smoothing the human-computer interaction aspect of any automated service at low cost. However, a lot of time and effort are required to prepare conversation. A botmaster has to prepare masses of expected questions and answers generally by himself. This is also true for a computer game dialogue preparation. This paper presents the design and implementation of a chatbot system that allows conversation about a particular topic to be constructed from existing pool of knowledge. Using Semi-CRF, we identify the role of each word in each sentence within the knowledge pool. With the role of each sub-sentence, patterns for questions and answers can be automatically generated. Our system was tested with 5 websites as knowledge pools. Our result shows that useful dialogues can be generated, saving time for conversation designers.