Network algorithms play a critical role in various applications, such as recommendations, diffusion maximization, and web search. In this paper, we focus on the fairness of such algorithms and in particular of PageRank. PageRank fairness refers to a fair allocation of the PageRank weights among the nodes. We consider the effect of the network structure on PageRank fairness. Concretely, we provide analytical formulas for computing the effect of edge additions on fairness and for the conditions that an edge must satisfy so that its addition improves fairness. We also provide analytical formulas for evaluating the role of existing edges in fairness. We use our findings to propose efficient linear time link recommendation algorithms for maximizing fairness, and we evaluate them on real datasets. Our approach can be seen as an effort towards making the network itself fairer as opposed to making fairer the network algorithms, or their outputs.

Temporal graphs represent relationships and interactions among entities over time, such as those occurring among users in social, transaction, and telecommunication networks. The analysis of their temporal structure help us understand, and predict the behavior of their entities. A typical analysis task in graph networks is the finding of all appearances of an input graph pattern query. Such appearances are called matches. In this paper, we are interested in finding all matches of an interaction pattern query within temporal graphs. To this end, we propose a hybrid approach that achieves effective filtering of potential matches based both on structure and time. Our approach exploits a graph representation where edges are ordered by time. We present experiments with real datasets that illustrate the efficiency of our approach.

Graphs offer a natural model for the relationships and interactions among entities, such as those occurring among users in social and cooperation networks, and proteins in biological networks. Since most such networks are dynamic, to capture their evolution over time, we assume a sequence of graph snapshots where each graph snapshot represents the state of the network at a different time instance. Given this sequence, we seek to find the top-$k$most durable matchesof an input graph pattern query, that is, the matches that exist for the longest period of time. The straightforward way to address this problem is to apply a state-of-the-art graph pattern matching algorithm at each snapshot and then aggregate the results. However, for large networks and long sequences, this approach is computationally expensive, since all matches have to be generated at each snapshot, including those appearing only once. We propose a new approach that uses a compact representation of the sequence of graph snapshots, appropriate time indexes to prune the search space, and strategies to determine the duration of the seeking matches. Finally, we present experiments with real datasets that illustrate the efficiency and effectiveness of our approach.

Efficiently detecting conversation threads from a pool of messages, such as social network chats, emails, comments to posts, news etc., is relevant for various applications, including Web Marketing, Information Retrieval and Digital Forensics. Existing approaches focus on text similarity using keywords as features that are strongly dependent on the dataset. Therefore, dealing with new corpora requires further costly analyses conducted by experts to find out new relevant features. This paper introduces a novel method to detect threads from any type of conversational texts overcoming the issue of previously determining specific features for each dataset. To automatically determine the relevant features of messages we map each message into a three dimensional representation based on its semantic content, the social interactions in terms of sender/recipients and its timestamp; then clustering is used to detect conversation threads. In addition, we propose a supervised approach to detect conversation threads that builds a classification model which combines the above extracted features for predicting whether a pair of messages belongs to the same thread or not. Our model harnesses the distance measure of a message to a cluster representing a thread to capture the probability that a message is part of that same thread. We present our experimental results on seven datasets, pertaining to different types of messages, and demonstrate the effectiveness of our method in the detection of conversation threads, clearly outperforming the state of the art and yielding an improvement of up to a 19%.

In this paper, we focus on labeled graphs that evolve over time. Given a sequence of graph snapshots representing the state of the graph at different time instants, we seek to find the most durable matches of an input graph pattern query, that is, the matches that exist for the longest period of time. The straightforward way to address this problem is by running a state-of-the-art graph pattern algorithm at each snapshot and aggregating the results. However, for large networks this approach is computationally expensive, since all matches have to be generated at each snapshot, including those appearing only once. We propose a new approach that uses a compact representation of the sequence of graph snapshots, appropriate time indexes to prune the search space and a threshold on the duration of the pattern to determine the search order. We also present experimental results using real datasets that illustrate the efficiency and effectiveness of our approach.

In this paper, we consider large graphs that evolve over time, such as graphs that model social networks. Given two instances of the graph at two points in time, we ask to identify the top pairs of nodes whose shortest path dis-tance has decreased the most. We call these pairs converg-ing. The straightforward way to address this problem is by computing the shortest path distances of all pairs at both instances and keeping the ones with the largest differences. Since for large networks this is computationally infeasible, we consider a budgeted version of the problem, where given a fixed budget of single-source shortest path computations, we seek to identify nodes that participate in as many con-verging pairs as possible. We evaluate a number of different approaches for our problem, that employ centrality-based, dispersion-based, and landmark-based distance estimation metrics. We also consider a classification-based approach that builds a classifier that combines the above features for predicting whether a node participates in one of the top con-verging pairs. We present experimental results using real-world datasets that show that we are able to identify the large majority of the top converging pairs on a very small budget.

Since most graphs evolve over time, it is useful to be able to query their history. We consider historical reachability queries that ask for the existence of a path in some time interval in the past, either in the whole duration of the interval (conjunctive queries), or in at least one time instant in the interval (disjunctive queries). We study both alternatives of storing the full transitive closure of the evolving graph and of performing an online traversal. Then, we propose an appropriate reachability index, termed TimeReach index, that exploits the fact that most real-world graphs contain large strongly connected components. Finally, we present an experimental evaluation of all approaches, for dierent graph sizes, historical query types and time granularities.