Yuzhe Zhang, Xianwei Xue, Xingyong Wu, Mengke Chen, Chen Liu, Xinran He, Run Shao, Feiran Liu, Huanmin Xu, Qiutong Pan, Haiwei Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Social platforms have paved the way in creating new, modern ways for users to communicate with each other. In recent years, multiple platforms have introduced ''Stories'' features, which enable broadcasting of ephemeral multimedia content. Specifically, ''Friend Stories,'' or Stories meant to be consumed by one's close friends, are a popular feature, promoting significant user-user interactions by allowing people to see (visually) what their friends and family are up to. A key challenge in surfacing Friend Stories for a given user, is in ranking over each viewing user's friends to efficiently prioritize and route limited user attention. In this work, we explore the novel problem of Friend Story Ranking from a graph representation learning perspective. More generally, our problem is a link ranking task, where inferences are made over existing links (relations), unlike common node or graph-based tasks, or link prediction tasks, where the goal is to make inferences about non-existing links. We propose ELR, an edge-contextual approach which carefully considers local graph structure, differences between local edge types and directionality, and rich edge attributes, building on the backbone of graph convolutions. ELR handles social sparsity challenges by considering and attending over neighboring nodes, and incorporating multiple edge types in local surrounding egonet structures. We validate ELR on two large country-level datasets with millions of users and tens of millions of links from Snapchat. ELR shows superior performance over alternatives by 8% and 5% error reduction measured by MSE and MAE correspondingly. Further generality, data efficiency and ablation experiments confirm the advantages of ELR.

In the well-studied Influence Maximization problem, the goal is to identify a set of k nodes in a social network whose joint influence on the network is maximized. A large body of recent work has justified research on Influence Maximization models and algorithms with their potential to create societal or economic value. However, in order to live up to this potential, the algorithms must be robust to large amounts of noise, for they require quantitative estimates of the influence, which individuals exert on each other; ground truth for such quantities is inaccessible, and even decent estimates are very difficult to obtain. We begin to address this concern formally. First, we exhibit simple inputs on which even very small estimation errors may mislead every algorithm into highly suboptimal solutions. Motivated by this observation, we propose the Perturbation Interval model as a framework to characterize the stability of Influence Maximization against noise in the inferred diffusion network. Analyzing the susceptibility of specific instances to estimation errors leads to a clean algorithmic question, which we term the Influence Difference Maximization problem. However, the objective function of Influence Difference Maximization is NP-hard to approximate within a factor of O ( n 1−ϵ) for any ϵ > 0. Given the infeasibility of diagnosing instability algorithmically, we focus on finding influential users robustly across multiple diffusion settings. We define a Robust Influence Maximization framework wherein an algorithm is presented with a set of influence functions. The algorithm’s goal is to identify a set of k nodes who are simultaneously influential for all influence functions, compared to the (function-specific) optimum solutions. We show strong approximation hardness results for this problem unless the algorithm gets to select at least a logarithmic factor more seeds than the optimum solution. However, when enough extra seeds may be selected, we show that techniques of Krause et al. can be used to approximate the optimum robust influence to within a factor of 1−1/ e . We evaluate this bicriteria approximation algorithm against natural heuristics on several real-world datasets. Our experiments indicate that the worst-case hardness does not necessarily translate into bad performance on real-world datasets; all algorithms perform fairly well.

Network Inference, i.e., discovering latent diffusion networks from observed cascades, has been studied extensively in recent years, leading to a series of excellent work. However, it has been observed that the accuracy of existing methods deteriorates significantly when the number of cascades are limited (compared with the large number of nodes), which is the norm in real world applications. Meanwhile, we are able to collect cascades on many different topics or over a long time period: the associated influence networks (either topic-specific or time-specific) are highly correlated while the number of cascade observations associated with each network is very limited. In this work, we propose a generative model, referred to as the MultiCascades model (MCM), to address the challenge of data scarcity by exploring the commonality between multiple related diffusion networks. MCM builds a hierarchical graphical model, where all the diffusion networks share the same network prior, e.g., the popular Stochastic Blockmodels or the latent space models. The parameters of the network priors can be effectively learned by gleaning evidence from a large number of inferred networks. In return, each individual network can be inferred more accurately thanks to the prior information. Furthermore, we develop efficient inference and learning algorithms so that MCM is scalable for practical applications. The results on both synthetic datasets and real-world datasets demonstrate that MCM infers both topic-specific and time-varying diffusion networks more accurately.

As both social network structure and strength of influence between individuals evolve constantly, it requires tracking the influential nodes under a dynamic setting. To address this problem, we explore the Influential Node Tracking (INT) problem as an extension to the traditional Influence Maximization problem (IM) under dynamic social networks. While the Influence Maximization problem aims at identifying a set of k nodes to maximize the joint influence under one static network, the INT problem focuses on tracking a set of influential nodes that keeps maximizing the influence as the network evolves. Utilizing the smoothness of the evolution of the network structure, we propose an efficient algorithm, Upper Bound Interchange Greedy (UBI) and a variant, UBI+. Instead of constructing the seed set from the ground, we start from the influential seed set we found previously and implement node replacement to improve the influence coverage. Furthermore, by using a fast update method by calculating the marginal gain of nodes, our algorithm can scale to dynamic social networks with millions of nodes. Empirical experiments on three real large-scale dynamic social networks show that our UBI and its variants, UBI+ achieves better performance in terms of both influence coverage and running time.

Uncertainty about models and data is ubiquitous in the computational social sciences, and it creates a need for robust social network algorithms, which can simultaneously provide guarantees across a spectrum of models and parameter settings. We begin an investigation into this broad domain by studying robust algorithms for the Influence Maximization problem, in which the goal is to identify a set of k nodes in a social network whose joint influence on the network is maximized. We define a Robust Influence Maximization framework wherein an algorithm is presented with a set of influence functions, typically derived from different influence models or different parameter settings for the same model. The different parameter settings could be derived from observed cascades on different topics, under different conditions, or at different times. The algorithm's goal is to identify a set of k nodes who are simultaneously influential for all influence functions, compared to the (function-specific) optimum solutions.

We study the problem of learning influence functions under incomplete observations of node activations. Incomplete observations are a major concern as most (online and real-world) social networks are not fully observable. We establish both proper and improper PAC learnability of influence functions under randomly missing observations. Proper PAC learnability under the Discrete-Time Linear Threshold (DLT) and Discrete-Time Independent Cascade (DIC) models is established by reducing incomplete observations to complete observations in a modified graph. Our improper PAC learnability result applies for the DLT and DIC models as well as the Continuous-Time Independent Cascade (CIC) model. It is based on a parametrization in terms of reachability features, and also gives rise to an efficient and practical heuristic. Experiments on synthetic and real-world datasets demonstrate the ability of our method to compensate even for a fairly large fraction of missing observations.

With increasing popularity of social media and social networks sites, analyzing the social networks offers great potential to shed light on human social structure and provides great marketing opportunities. Usually, social network analysis starts with extracting or learning the social network and the associated parameters. Contrary to other analytical tasks, this step is highly non-trivial due to amorphous nature of social ties and the challenges of noisy and incomplete observations. My research focuses on improving accuracy in inferring the network as well as analyzing the consequences when the extracted network is noisy or erroneous. To be more precise, I propose to study the following two questions with a special focus on analyzing diffusion behaviors: (1) How to utilize special properties of social networks to improve accuracy of the extracted network under noisy and missing data; (2) How to characterize the impact of noise in the inferred network and carry out robust analysis and optimization.

Topic models have achieved significant successes in analyzing large-scale text corpus. In practical applications, we are always confronted with the challenge of model selection, i.e., how to appropriately set the number of topics. Following the recent advances in topic models via tensor decomposition, we make a first attempt to provide theoretical analysis on model selection in latent Dirichlet allocation. With mild conditions, we derive the upper bound and lower bound on the number of topics given a text collection of finite size. Experimental results demonstrate that our bounds are correct and tight. Furthermore, using Gaussian mixture model as an example, we show that our methodology can be easily generalized to model selection analysis in other latent models.

Understanding the diffusion of information in social network and social media requires modeling the text diffusion process. In this work, we develop the HawkesTopic model (HTM) for analyzing text-based cascades, such as "retweeting a post" or "publishing a follow-up blog post". HTM combines Hawkes processes and topic modeling to simultaneously reason about the information diffusion pathways and the topics characterizing the observed textual information. We show how to jointly infer them with a mean-field variational inference algorithm and validate our approach on both synthetic and real-world data sets, including a news media dataset for modeling information diffusion, and an ArXiv publication dataset for modeling scientific influence. The results show that HTM is significantly more accurate than several baselines for both tasks.

Real world marketing campaign utilizing the word-of-mouth effect usually lasts a long time, where multiple sets of influential users need to be mined and targeted at different times to fully utilize the power of viral marketing. As both social network structure and strength of influence between individuals evolve constantly, it requires to track the influential nodes under a dynamic setting. To address the above problem, we explore the Influential Node Tracking (INT) problem as an extension to the traditional Influence Maximization problem under dynamic social networks. While Influence Maximization problem aims at identifying a set of k nodes to maximize the joint influence under one static network, INT problem focuses on tracking a set of influential nodes that keeps maximizing the influence as the network evolves. Utilizing the smoothness of the evolution of the network structure, we propose an efficient algorithm, Upper Bound Interchange Greedy (UBI) to solve the INT problem. Instead of constructing the seed set from the ground, we start from the influential seed set we find previously and implement node replacement to improve the influence coverage. Furthermore, by using a fast update method to maintain an upper bound on the node replacing gain, our algorithm can scale to dynamic social networks with millions of nodes. Empirical experiments on three real large-scale dynamic social networks show that our UBI algorithm achieves better performance in terms of both influence coverage and running time.

Traditional anomaly detection on social media mostly focuses on individual point anomalies while anomalous phenomena usually occur in groups. Therefore, it is valuable to study the collective behavior of individuals and detect group anomalies. Existing group anomaly detection approaches rely on the assumption that the groups are known, which can hardly be true in real world social media applications. In this article, we take a generative approach by proposing a hierarchical Bayes model: Group Latent Anomaly Detection (GLAD) model. GLAD takes both pairwise and point-wise data as input, automatically infers the groups and detects group anomalies simultaneously. To account for the dynamic properties of the social media data, we further generalize GLAD to its dynamic extension d-GLAD. We conduct extensive experiments to evaluate our models on both synthetic and real world datasets. The empirical results demonstrate that our approach is effective and robust in discovering latent groups and detecting group anomalies.

The present article serves as an erratum to our paper of the same title, which was presented and published in the KDD 2014 conference. In that article, we claimed falsely that the objective function defined in Section 1.4 is non-monotone submodular. We are deeply indebted to Debmalya Mandal, Jean Pouget-Abadie and Yaron Singer for bringing to our attention a counter-example to that claim.

Traditional anomaly detection on social media mostly focuses on individual point anomalies while anomalous phenomena usually occur in groups. Therefore it is valuable to study the collective behavior of individuals and detect group anomalies. Existing group anomaly detection approaches rely on the assumption that the groups are known, which can hardly be true in real world social media applications. In this paper, we take a generative approach by proposing a hierarchical Bayes model: Group Latent Anomaly Detection (GLAD) model. GLAD takes both pair-wise and point-wise data as input, automatically infers the groups and detects group anomalies simultaneously. To account for the dynamic properties of the social media data, we further generalize GLAD to its dynamic extension d-GLAD. We conduct extensive experiments to evaluate our models on both synthetic and real world datasets. The empirical results demonstrate that our approach is effective and robust in discovering latent groups and detecting group anomalies.

In many real-world situations, different and often opposite opinions, innovations, or products are competing with one another for their social influence in a networked society. In this paper, we study competitive influence propagation in social networks under the competitive linear threshold (CLT) model, an extension to the classic linear threshold model. Under the CLT model, we focus on the problem that one entity tries to block the influence propagation of its competing entity as much as possible by strategically selecting a number of seed nodes that could initiate its own influence propagation. We call this problem the influence blocking maximization (IBM) problem. We prove that the objective function of IBM in the CLT model is submodular, and thus a greedy algorithm could achieve 1 — 1/e approximation ratio. However, the greedy algorithm requires Monte-Carlo simulations of competitive influence propagation, which makes the algorithm not efficient. We design an efficient algorithm CLDAG, which utilizes the properties of the CLT model, to address this issue. We conduct extensive simulations of CLDAG, the greedy algorithm, and other baseline algorithms on real-world and synthetic datasets. Our results show that CLDAG is able to provide best accuracy in par with the greedy algorithm and often better than other algorithms, while it is two orders of magnitude faster than the greedy algorithm.