Online social networks nowadays enjoy their worldwide prosperity, as they have revolutionized the way for people to discover, to share, and to distribute information. With millions of registered users and the proliferation of user-generated contents, the social networks be-come "giants", likely eligible to carry on any research tasks. How-ever, the giants do have their Achilles Heel: extreme data sparsity. Compared with the massive data over the whole collection, individ-ual posting documents, (e.g., a microblog less than 140 characters), seem to be too sparse to make a difference under various research scenarios, while actually they are different. In this paper we pro-pose to tackle the Achilles Heel of social networks by smoothing the language model via influence propagation. We formulate a so-cialized factor graph model, which utilizes both the textual corre-lations between document pairs and the socialized augmentation networks behind the documents, such as user relationships and so-cial interactions. These factors are modeled as attributes and de-pendencies among documents and their corresponding users. An efficient algorithm is designed to learn the proposed factor graph model. Finally we propagate term counts to smooth documents based on the estimated influence. Experimental results on Twitter and Weibo datasets validate the effectiveness of the proposed mod-el. By leveraging the smoothed language model with social factors, our approach obtains significant improvement over several alterna-tive methods on both intrinsic and extrinsic evaluations measured in terms of perplexity, nDCG and MAP results.