Detecting Information Need In Large Scale Social Media Content

In this paper, we present the idea of detecting Information Need in large social media network. Information need is the need from people to locate and obtain specific information. Social medias can benefit the users by successfully detecting such needs. We formulate this problem as a classification problem, adopting multiple sources of feature and perform linear learning. We propose a distributed algorithm for large-scale linear classification using the alternating direction method of multipliers coupled with dual coordinate descent. Our experiments show that (1) the proposed distributed algorithm outperforms state-of-art solvers on large data sets in a distributed system (2) the proposed solution is effective in our Twitter data set.