Adhitya Bhawiyuga, M. Ali Fauzi, Eko Sakti Pramukantoro, Widhi Yahya
In order to provide an excellent service, the seller in e-commerce world is required to actively involves in communication with its customer. Nevertheless, in several condition e.g. in vacation or during a rest, the seller might be unable to communicate with his/her customer. While employing a customer service can be a partial solution, it may involves additional costs for paying customer service persons. In this paper, we propose the design and implementation of e-commerce chatbot system which provides an automatic response to the incoming customer-to-seller question. In general, the proposed system consists of two main agents : communication and intelligent part. In order to get the question message sent by the customer, the communication agent periodically performs a request to Telegram server using a standard HTTP protocol. Upon reception, it forwards that question to intelligent agent which then find the closest instance in predefined question-answer corpora. Notice that, we utilize the Levenstein distance to measure the difference between a submitted question with that of in predefined question-answer corpora. Once an closest instance is selected, the intelligent agent forward the answer to communication agent which then send the answer back to the sender through Telegram chat service. From usability and performance testing result, the proposed system can deliver the automatic answer in less than 5 seconds with relatively good matching accuracy. © 2017 IEEE.
Faculty of Computer Science, University of Brawijaya, Malang, Indonesia
