Automated content based short text classification for filtering undesired posts on Facebook

Abstract

Online Social Networking (OSN) sites are always helpful for being socialized and to get exposed to a social environment. But, privacy and prevention of undesired posts on user wall is the only problem of biggest concern. User should have the ability to control the message posted on their own private wall to avoid undesirable contents to be displayed. The existing OSN sites have very little support regarding this problem. For example, Facebook filters messages on the basis of identity of sender i.e. only friend, friend of friend or group of friends can post any message; no content based preferences are supported. Taking this fact into consideration, the proposed work contributes to address such problem through a machine learning based soft classifier for labeling messages in support of contents of message. This work experimentally evaluates an automated scheme to filter out unwanted messages posted on Facebook walls by assigning a set of categories with each short text message based on its contents.