| Nowadays, large scale functional website operation process now becomes more and more complex and have more and more functions. In ordering to provide guidance to users, solve problems that users face when using the website, there are three kinds of method. The first method is to set up specialized’Frequently Asked Questions’web pages and spare specialized personnel to maintain the pages. For websites which have more complicated functions, they sometimes have to use the second method which is to spare specialized personnel for customer service. This traditional way takes a lot of cost of manpower for the company and FQA and custom service seem not to be a convenient way for user to find an answer. The third method is the question answer system. This project belonging to question answer system, is named ECSA(eBay Customer Service Assistant), aiming to provide a unified interface for website users and solve problems users encounter. The system can craw question related to the website and corresponding answer from question answering websites like Yahoo! Answer, filter customer service QApairs from these QA pairs by a SVM based classifier and answer users’question.The system is built using Play Framework. The system trains the model by pre-classified samples using Libsvm and does continuous classification through this model. To analyze English text we use Stanford library which can mark part-of-speech of sentences. Mysql is used as formal database and we also apply revert index to accelerate searching speed.The system is designed as a layered architecture. Each layer has the characteristics of modular and loosely coupled. The bottom layer manages data obtained by a crawler, types of data include files, databases and reverted index. The data processing layer contains a small manual labeling system and a SVM based classifier. The business process layer is mainly responsible for interactions with users. This layer contains lexial analyzer and all business logics. |
PDF Full Text Request