@utrust_user

Comment classification for Internet auction platforms

, , and . Local Proceedings of 13th East-European Conference, ADBIS 2009, page 374-384. JUMI Pubbbblishing House Ltd., (2009)

Abstract

Reducing Internet auction fraud is one of the greatest challenges in today’s electronic market. Most of the electronic auction platforms use only simple reputation system that can be easily manipulated. Although reputation systems can be used to detect frauds, they provide little detailed information about the fraud itself except user comments. Significant experience is required to understand every aspect of Internet auctions. Despite many help pages and tutorials provided by auction platforms, in most cases it is not easy to teach users how to protect themselves from Internet fraud. When all negative comments are treated equally by reputation algorithms, it is impossible to distinguish between malicious behavior and accidental mistakes. Our goal is to design automatic comment classification methods that will allow a meaningful distinction of different types of negative and neutral comments. We have developed a hierarchical model of user behavior in Internet auctions (separately for buyers and sellers). By checking the frequency of occurrence and the significance of reported transaction problems we have created simple classification method to detect potential threats related to users’ transactions. We have also proposed method of rating complaints against sellers and buyers that can be used to modify the Internet auction reputation algorithms.

Links and resources

Tags

community

  • @utrust_user
  • @dblp
@utrust_user's tags highlighted