부산대학교 도서관

In any business, a humongous amount of data is generated within each fraction of a second by each and every software application. However, processing this voluminous data is a tedious task especially when the data is in textual form. In this scenario, Natural Language Processing has contributed majorly in this area of study. In NLP, keyword extraction plays a pivotal role in text processing which helps the readers to determine whether to read a document or a webpage. The system designed in this paper computed extractive text summarization using Graph-based technique and TextRank algorithm on conversation data between the user and the service chatbots which in fact an offline conversation. This summary is then consumed by a classification module which is trained using Naive Bayes classifier to evaluate in which of the three categories the conversation falls into: 1. Help 2.Complaint 3. FakeCustomer. This system can be utilized by various companies such as online shopping websites, software companies to determine in which aspect immediate attention is required. The system is also experimented using different thresholds for determining the length of summary produced and its corresponding accuracy.