Social clues powered, personalized software engineering messages classification

Software Engineering is continuously flourished with model, methodologies, tools, and best practices. Only by investing a great deal of time can one keep up with the cutting-edge knowledge. Active and numerous communications on Twitter make it ideal for the task. Even though, content-based information extraction is not quite successful due to the message diversity and implicit user interests. To improve the classification accuracy, we proposed the classification architecture that embraces social clues which consist of the user profile, and user impact. Three features of the messages, the content feature, the user feature and the community feature, are defined according to the social clues and are used to classify the message. The new architecture yield 35 percents of the accuracy improvement from sole content-based classification. By exploiting the social clues, we push the Twitter message classification to the next level.