LogRank: Summarizing Social Activity Logs

Online Social Networks (OSNs) allow users to create and share content (e.g., posts, status updates, comments) in real-time. These activities are collected in an activity log, (e.g. Facebook Wall, Google+ Stream, etc.) on the user’s social network profile. With time, the activity logs of users, which record the sequences of social activities, become too long and consequently hard to view and navigate. To alleviate this cluttering, it is useful to select a small subset of the social activities within the specified timeperiod as representative, i.e., as summary, for this time-period. In this paper, we study the novel problem of social activity log summarization. We propose LogRank, a novel and principled algorithm to select activities that satisfy three desirable criteria: First, activities must be important for the user. Second, they must be diverse in terms of topic, e.g., cover several of the major topics in the activity log. Third, they should be time-dispersed, that is, be spread across the specified time range of the activity log. LogRank operates on an appropriately augmented social interaction graph and employs random-walk techniques to holistically balance all three criteria. We evaluate LogRank and its variants on a real dataset from the Google+ social network and show that they outperform baseline approaches.