Selected Publications
This is a selected list of publications.
For others, see my DBLP entry or Google Scholar profile.
Social Media Mining
An Introduction
Reza Zafarani
Mohammad Ali Abbasi
Huan Liu
A Textbook by
Cambridge University Press, 2014
Download
An Introduction
Reza Zafarani
Mohammad Ali Abbasi
Huan Liu
A Textbook by
Cambridge University Press, 2014
Download
Filter by type:
Sort by year:
Harnessing the Power of Noise: A Survey of Techniques and Applications
Journal PaperACM Computing Surveys (2026)
Abstract
Noise, traditionally considered a nuisance in computational systems, is reconsidered for its unexpected and counter-intuitive benefits across a wide spectrum of domains, including nonlinear information processing, signal processing, image processing, machine learning, network science, and natural language processing. Through a comprehensive review of both historical and contemporary research, this survey presents a dual perspective on noise, acknowledging its potential to both disrupt and enhance performance. Particularly, we highlight how noise-enhanced training strategies can lead to models that better generalize from noisy data, positioning noise not just as a challenge to overcome but as a strategic tool for improvement. This work calls for a shift in how we perceive noise, proposing that it can be a spark for innovation and advancement in the information era.
When Does Demographic Information Help? Data and Modeling Regimes for Perspective-Aware Hate Speech Detection
PreprintarXiv:2605.27313
Abstract
Demographic information is often used to model annotator perspectives in subjective tasks such as hate speech detection, but its benefit is inconsistent: it improves performance in some settings and behaves as noise in others. This paper asks when demographic features help. We analyze demographic gain as a function of both data split properties and modeling frameworks. For data splits, we measure annotator disagreement, namely how often annotators assign different labels to the same example, along with training size and train-test demographic coverage. We find that demographic gains concentrate in regimes with low training disagreement, high test disagreement, fine-grained ambiguity measurement, sufficient training data, and greater demographic overlap. Motivated by these regimes, we introduce a gated demographic residual model that treats demographics as a selective adjustment to text-only predictions. Experiments on MHS and POPQUORN show that this design is effective, especially on high disagreement or low confidence examples. Overall, our results suggest that demographics should not be assumed useful by default; their value depends jointly on the data regime and the modeling framework.
Abstract
The propagation structure of fake news has been shown to be an important cue for detecting it; yet, existing propagation-based fake news detection methods have mainly relied on ad hoc topological features, and a unified view of cascade patterns is still lacking. To address this, we study news propagation from a spectral view by connecting graph spectra to propagation-related structural properties through rigorous spectral bounds. In particular, we introduce several new bounds and integrate them with existing ones into a unified spectral representation of information propagation. We then use these spectral bounds for downstream classification and design a discrete structural optimization framework to interpret learned propagation patterns. For efficient optimization, we rely on a first-order perturbation approximation and consider both score-guided and bound-guided objectives. Experiments on real-world data reveal meaningful spectral differences between fake and real news, competitive classification performance from spectral bounds, and interpretable evolution trajectories from structural optimization. The findings demonstrate the value of spectral analysis for understanding and modeling news propagation.
Analyzing Memory Effects in Large Language Models through the Lens of Cognitive Psychology
PreprintarXiv:2509.17138
Abstract
Large Language Models (LLMs) exhibit memory-like behaviors that resemble aspects of human cognition. In this work, we examine such behaviors through the lens of cognitive psychology, drawing parallels between LLM outputs and classical findings on human memory including primacy/recency effects, interference, and forgetting curves. By bridging machine learning and cognitive science, we offer a framework for interpreting LLM behavior and discuss implications for evaluation, reliability, and the design of more human-aligned AI systems.
Seeing Hate Differently: Hate Subspace Modeling for Culture-Aware Hate Speech Detection
PreprintarXiv:2510.13837
Abstract
Hate speech detection faces significant challenges when applied across cultural contexts, as expressions considered hateful in one culture may be acceptable in another. In this work, we propose a culture-aware framework for hate speech detection that models distinct "hate subspaces" reflecting how different cultural groups perceive and express hate. Our approach moves beyond one-size-fits-all classification and captures the nuanced, culturally dependent nature of harmful language online.
Abstract
While memes are often humorous, they are frequently used to disseminate hate, causing serious harm to individuals and society. Current approaches to hateful meme detection mainly rely on pre-trained language models. However, less focus has been dedicated to what makes a meme hateful. Drawing on insights from philosophy and psychology, we argue that hateful memes are characterized by two essential features: a presupposed context and the expression of false claims. To capture presupposed context, we develop PCM for modeling contextual information across modalities. To detect false claims, we introduce the FACT module, which integrates external knowledge and harnesses cross-modal reference graphs.
Is Less Really More? Fake News Detection with Limited Information
Journal PaperACM SIGKDD Explorations Newsletter 27.1 (2025), pp. 20--31
Abstract
The threat that online fake news and misinformation pose to democracy, justice, public confidence, and especially to vulnerable populations, has led to a sharp increase in the need for fake news detection and intervention. Whether multi-modal or pure text-based, most fake news detection methods depend on textual analysis of entire articles. However, these methods face challenges including large training data requirements, sensitivity to topic changes, and the difficulty of encoding lengthy articles. This paper investigates whether effective fake news detection is feasible with limited textual information, focusing only on article source domains rather than full content.
Representing Higher-Order Networks with Spectral Moments
Conference PaperProceedings of the 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD), Sydney, Australia (2025)
Abstract
The spectral properties of traditional (dyadic) graphs, where an edge connects exactly two vertices, are widely utilized in different applications. These spectral properties are closely connected to the structural properties of dyadic graphs. We generalize such connections and characterize higher-order networks by their spectral information. We first split the higher-order graphs by their orders into several uniform hypergraphs. For each uniform hypergraph, we extract the corresponding spectral information from the transition matrices of carefully designed random walks. From each spectrum, we compute the first few spectral moments and use all such spectral moments across different orders as the higher-order graph representation. We will show that these moments not only clearly indicate the return probabilities of random walks but are also closely related to various higher-order network properties such as degree distribution and clustering coefficient. Extensive experiments show the utility of this new representation in various settings.
Higher-Order Networks Representation and Learning: A Survey
Journal PaperACM SIGKDD Explorations Newsletter 26.1 (2024), pp. 1--18
Abstract
Network data has become widespread, larger, and more complex over the years. Traditional network data is dyadic, capturing the relations among pairs of entities. With the need to model interactions among more than two entities, significant research has focused on higher-order networks and ways to represent, analyze, and learn from them. There are two main directions to studying higher-order networks. One direction has focused on capturing higher-order patterns in traditional (dyadic) graphs by changing the basic unit of study from nodes to small frequently observed subgraphs, called motifs. As most existing network data comes in the form of pairwise dyadic relationships, studying higher-order structures within such graphs may uncover new insights. The second direction aims to directly model higher-order interactions using new and more complex representations such as simplicial complexes or hypergraphs. Some of these models have long been proposed, but improvements in computational power and the advent of new computational techniques have increased their popularity. Our goal in this paper is to provide a succinct yet comprehensive summary of the advanced higher-order network analysis techniques. We provide a systematic review of its foundations and algorithms, along with use cases and applications of higher-order networks in various scientific domains.
Abstract
The rapid rise of large language models has enabled the automated generation of fake news at scale, with content that is increasingly difficult to distinguish from human-written text. In this work, we study AI-generated fake news across multiple domains and develop techniques to detect and characterize such content. We analyze the linguistic and stylistic signatures of machine-generated misinformation, evaluate the performance of detection methods under domain shift, and discuss the implications for combating automated disinformation campaigns.
Semi-Supervised Graph Ultra-Sparsifier Using Reweighted 
Optimization
Conference PaperProceedings of the 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2023)
Abstract
Graph representation learning with the family of graph convolution networks (GCN) provides powerful tools for prediction on graphs. As graphs grow with more edges, the GCN family suffers from sub-optimal generalization performance due to task-irrelevant connections. Recent studies solve this problem by using graph sparsification in neural networks. However, graph sparsification cannot generate ultra-sparse graphs while simultaneously maintaining the performance of the GCN family. To address this problem, we propose Graph Ultra-sparsifier, a semi-supervised graph sparsifier with dynamically-updated regularization terms based on the graph convolution. The graph ultra-sparsifier can generate ultra-sparse graphs while maintaining the performance of the GCN family with the ultra-sparse graphs as inputs. In the experiments, when compared to the state-of-the-art graph sparsifiers, our graph ultra-sparsifier generates ultra-sparse graphs and these ultra-sparse graphs can be used as inputs to maintain the performance of GCN and its variants in node classification tasks.
Abstract
We propose the hierarchical recursive neural network (HERO) to predict fake news by learning its linguistic style, which is distinguishable from the truth, as psychological theories reveal. We first generate the hierarchical linguistic tree of news documents; by doing so, we translate each news document's linguistic style into its writer's usage of words and how these words are recursively structured as phrases, sentences, paragraphs, and, ultimately, the document. By integrating the hierarchical linguistic tree with the neural network, the proposed method learns and classifies the representation of news documents by capturing their locally sequential and globally recursive structures that are linguistically meaningful. It is the first work offering the hierarchical linguistic tree and the neural network preserving the tree information to our best knowledge. Experimental results based on public real-world datasets demonstrate the proposed method's effectiveness, which can outperform state-of-the-art techniques in classifying short and long news documents. We also examine the differential linguistic style of fake news and the truth and observe some patterns of fake news. The code and data have been publicly available.
The Advantages of Adding Noise
TutorialCompanion Proceedings of The Web Conference 2023 (WWW) (2023)
Abstract
Noise is often seen as unwanted signal while it has been shown beneficial in many information processing systems and algorithms. Noise enhancement has been utilized in many biological and physical systems, machine learning methods, and deep learning techniques in order to improve efficiency and performance. This tutorial presents (1) the different types of noise; (2) noise applications; (3) noise-enhanced processing systems; (4) noise-enhanced learning methods; and (5) noise injection methods in network science.
Interpretable Network Representations
TutorialCompanion Proceedings of The Web Conference 2022 (WWW) (2022)
Abstract
Networks (or interchangeably graphs) have been ubiquitous across the globe and within science and engineering: social networks, collaboration networks, protein-protein interaction networks, infrastructure networks, among many others. Machine learning on graphs, especially network representation learning, has shown remarkable performance in tasks related to graphs, such as node/graph classification, graph clustering, and link prediction. These tasks are closely related to the Web applications, especially social network analysis and recommendation systems. For example, node classification and graph clustering are widely used for studies on community detection, and link prediction plays a vital role in friend or item recommendation. Like performance, it is equally crucial for individuals to understand the behavior of machine learning models and be able to explain how these models arrive at a certain decision. Such needs have motivated many studies on interpretability in machine learning. Specifically, for social network analysis, we may need to know the reasons why certain users (or groups) are classified or clustered together by the machine learning models, or why a friend recommendation system considers some users similar so that they are recommended to connect with each other. Under such circumstances, an interpretable network representation is necessary and it should carry the graph information to a level understandable by humans. In this tutorial, we will (1) define interpretability and go over its definitions within different contexts in studies of networks; (2) review and summarize various interpretable network representations; (3) discuss connections to network embedding, graph summarization, and network visualization methods; (4) discuss explainability in Graph Neural Networks, as such techniques are often perceived to have limited interpretability; and (5) highlight the open research problems and future research directions. The tutorial is designed for researchers, graduate students, and practitioners in areas such as graph mining, machine learning on graphs, and machine learning interpretability. Few prerequisites are required for The Web Conferenc participants to attend.
Exploiting Cross-Order Patterns and Link Prediction in Higher-Order Networks
Conference PaperProceedings of the IEEE International Conference on Data Mining Workshops (ICDMW) (2022)
Abstract
With the demand to model the relationships among three or more entities, higher-order networks are now more widespread across various domains. Relationships such as multiauthor collaborations, co-appearance of keywords, and copurchases can be naturally modeled as higher-order networks. However, due to (1) computational complexity and (2) insufficient higher-order data, exploring higher-order networks is often limited to order-3 motifs (or triangles). To address these problems, we explore and quantify similarites among various network orders. Our goal is to build relationships between different network orders and to solve higher-order problems using lowerorder information. Similarities between different orders are not comparable directly. Hence, we introduce a set of general crossorder similarities, and a measure: subedge rate. Our experiments on multiple real-world datasets demonstrate that most higherorder networks have considerable consistency as we move from higher-orders to lower-orders. Utilizing this discovery, we develop a new cross-order framework for higher-order link prediction method. These methods can predict higher-order links from lower-order edges, which cannot be attained by current higherorder methods that rely on data from a single order.
"This is Fake! Shared it by Mistake": Assessing the Intent of Fake News Spreaders
Conference PaperProceedings of The Web Conference 2022 (WWW) (2022)
Abstract
Individuals can be misled by fake news and spread it unintentionally without knowing that it is false. This phenomenon has been frequently observed but has not been investigated. Our aim in this work is to assess the intent of fake news spreaders. To distinguish between intentional versus unintentional spreading, we study the psychological interpretations behind unintentional spreading. With this foundation, we then propose an influence graph, using which we assess the intent of fake news spreaders. Our extensive experiments show that the assessed intent can help significantly differentiate between intentional and unintentional fake news spreaders. Furthermore, the estimated intent can significantly improve the current techniques that detect fake news. To our best knowledge, this is the first work to model individuals' intent in fake news spreading.
Noise Enhancement: Techniques and Applications
TutorialProceedings of the 2022 SIAM International Conference on Data Mining (SDM) (2022)
Abstract
Noise is often seen as unwanted signal while it has been shown beneficial in many information processing systems and algorithms. Noise enhancement has been utilized in many biological and physical systems, machine learning methods, and deep learning techniques in order to improve efficiency and performance. This tutorial presents (1) the different types of noise; (2) noise applications; (3) noise-enhanced processing systems; (4) noise-enhanced learning methods; and (5) noise injection methods in network science.
A Spectral Representation of Networks: The Path of Subgraphs
Conference PaperProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) (2022)
Abstract
Network representation learning has played a critical role in studying networks. One way to study a graph is to focus on its spectrum, i.e., the eigenvalue distribution of its associated matrices. Recent advancements in spectral graph theory show that spectral moments of a network can be used to capture the network structure and various graph properties. However, sometimes networks with different structures or sizes can have the same or similar spectral moments, not to mention the existence of the cospectral graphs. To address such problems, we propose a 3D network representation that relies on the spectral information of subgraphs: the Spectral Path, a path connecting the spectral moments of the network and those of its subgraphs of different sizes. We show that the spectral path is interpretable and can capture relationship between a network and its subgraphs, for which we present a theoretical foundation. We demonstrate the effectiveness of the spectral path in applications such as network visualization and network identification.
From Fake News to #FakeNews: Mining Direct and Indirect Relationships among Hashtags for Fake News Detection
PreprintarXiv:2211.11113
Abstract
The COVID-19 pandemic has gained worldwide attention and allowed fake news, such as ``COVID-19 is the flu,'' to spread quickly and widely on social media. Combating this coronavirus infodemic demands effective methods to detect fake news. To this end, we propose a method to infer news credibility from hashtags involved in news dissemination on social media, motivated by the tight connection between hashtags and news credibility observed in our empirical analyses. We first introduce a new graph that captures all (direct and \textit{indirect\}) relationships among hashtags. Then, a language-independent semi-supervised algorithm is developed to predict fake news based on this constructed graph. This study first investigates the indirect relationship among hashtags; the proposed approach can be extended to any homogeneous graph to capture a comprehensive relationship among nodes. Language independence opens the proposed method to multilingual fake news detection. Experiments conducted on two real-world datasets demonstrate the effectiveness of our approach in identifying fake news, especially at an \textit{early\} stage of propagation.
A Spectral Measure for Network Robustness: Assessment, Design, and Evolution
Conference PaperProceedings of the IEEE International Conference on Knowledge Graph (ICKG) (2022)
Abstract
A robust system should perform well under random failures or targeted attacks, and networks have been widely used to model the underlying structure of complex systems such as communication, infrastructure, and transportation networks. Hence, network robustness becomes critical to understanding system robustness. In this paper, we propose a spectral measure for network robustness: the second spectral moment m2 of the network. Our results show that a smaller second spectral moment m2 indicates a more robust network. We demonstrate both theoretically and with extensive empirical studies that the second spectral moment can help (1) capture various traditional measures of network robustness; (2) assess the robustness of networks; (3) design networks with controlled robustness; and (4) study how complex networked systems (e.g., power systems) behave under cascading failures.
AdverSparse: An Adversarial Attack Framework for Deep Spatial-Temporal Graph Neural Networks
Conference PaperProceedings of the 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2022)
Abstract
Spatial-temporal graph have been widely observed in various domains such as neuroscience, climate research, and transportation engineering. The state-of-the-art models of spatial-temporal graphs rely on Graph Neural Networks (GNNs) to obtain explicit representations for such networks and to discover hidden spatial dependencies in them, leading to superior performance in various tasks. In this paper, we propose a sparse adversarial attack framework AdverSparse to illustrate that when only a few key connections are removed in such graphs, hidden spatial dependencies learned by such spatial-temporal models are significantly impacted, leading to various issues such as increasing prediction errors. We formulate the adversarial attack on such models as an optimization problem and solve it by the Alternating Direction Method of Multipliers (ADMM). Experiments show that AdverSparse can find and remove key connections in these graphs, leading to malfunctioning models, even in models capable of learning hidden spatial dependencies.
Abstract
COVID-19 has impacted all lives. To maintain social distancing and avoiding exposure, works and lives have gradually moved online. Under this trend, social media usage to obtain COVID-19 news has increased. Alas, misinformation on COVID-19 is frequently spread on social media. In this work, we develop CHECKED, the first Chinese dataset on COVID-19 misinformation. CHECKED provides ground-truth on credibility, carefully obtained by ensuring the specific sources are used. CHECKED includes microblogs related to COVID-19, identified by using a specific list of keywords, covering a total 2120 microblogs published from December 2019 to August 2020. The dataset contains a rich set of multimedia information for each microblog including ground-truth label, textual, visual, response, and social network information. We hope that CHECKED can facilitate studies that target misinformation on coronavirus. The dataset is available at this https URL with measures of protecting users' privacy.
Graph Sparsification with Graph Convolutional Networks
Journal PaperInternational Journal of Data Science and Analytics (2021)
Abstract
Graphs are ubiquitous across the globe and within science and engineering. Some powerful classifiers are proposed to classify nodes in graphs, such as Graph Convolutional Networks (GCNs). However, as graphs are growing in size, node classification on large graphs can be space and time consuming due to using whole graphs. Hence, some questions are raised, particularly, whether one can prune some of the edges of a graph while maintaining prediction performance for node classification, or train classifiers on specific subgraphs instead of a whole graph with limited performance loss in node classification. To address these questions, we propose Sparsified Graph Convolutional Network (SGCN), a neural network graph sparsifier that sparsifies a graph by pruning some edges. We formulate sparsification as an optimization problem and solve it by an Alternating Direction Method of Multipliers (ADMM). The experiment illustrates that SGCN can identify highly effective subgraphs for node classification in GCN compared to other sparsifiers such as Random Pruning, Spectral Sparsifier and DropEdge. We also show that sparsified graphs provided by SGCN can be inputs to GCN, which leads to better or comparable node classification performance with that of original graphs in GCN, DeepWalk, GraphSAGE, and GAT. We provide insights on why SGCN performs well by analyzing its performance from the view of a low-pass filter.
Noise-Enhanced Unsupervised Link Prediction
Conference PaperProceedings of the 25th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD) (2021)
Abstract
Link prediction has attracted attention from multiple research areas. Although several — mostly unsupervised — link prediction methods have been proposed, improving them is still under study. In several fields of science, noise is used as an advantage to improve information processing, inspiring us to also investigate noise enhancement in link prediction. In this research, we study link prediction from a data preprocessing point of view by introducing a noise-enhanced link prediction framework that improves the links predicted by current link prediction heuristics. The framework proposes three noise methods to help predict better links. Theoretical explanation and extensive experiments on synthetic and real-world datasets show that our framework helps improve current link prediction methods.
SAFE: Similarity-Aware Multi-Modal Fake News Detection
Conference PaperProceedings of the 24th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD) (2020)
Abstract
Effective detection of ``fake news'' has recently attracted significant attention. Current studies have made significant contributions to predicting fake news with less focus on exploiting the relationship (similarity) between the textual and visual information in news articles. Attaching importance to such similarity helps identify fake news stories that, for example, attempt to use irrelevant images to attract readers' attention. In this work, we propose a $\mathsf{S\}$imilarity-$\mathsf{A\}$ware $\mathsf{F\}$ak$\mathsf{E\}$ news detection method ($\mathsf{SAFE\}$) which investigates multi-modal (textual and visual) information of news articles. First, neural networks are adopted to separately extract textual and visual features for news representation. We further investigate the relationship between the extracted features across modalities. Such representations of news textual and visual information along with their relationship are jointly learned and used to predict fake news. The proposed method facilitates recognizing the falsity of news articles based on their text, images, or their ``mismatches.'' We conduct extensive experiments on large-scale real-world data, which demonstrate the effectiveness of the proposed method.
WebShapes: Network Visualization with 3D Shapes
Conference PaperProceedings of the 13th ACM International Conference on Web Search and Data Mining (WSDM) (2020)
Abstract
Network visualization has played a critical role in graph analysis, as it not only presents a big picture of a network but also helps reveal the structural information of a network. The most popular visual representation of networks is the node-link diagram. However, visualizing a large network with the node-link diagram can be challenging due to the difficulty in obtaining an optimal graph layout. To address this challenge, a recent advancement in network representation: network shape, allows one to compactly represent a network and its subgraphs with the distribution of their embeddings. Inspired by this research, we have designed a web platform WebShapes that enables researchers and practitioners to visualize their network data as customized 3D shapes (<a href="http://b.link/webshapes">http://b.link/webshapes</a>)Furthermore, we provide a case study on real-world networks to explore the sensitivity of network shapes to different graph sampling, embedding, and fitting methods, and we show examples of understanding networks through their network shapes.
Sentiment Paradoxes in Social Networks: Why Your Friends are More Positive Than You?
Conference PaperProceedings of the 14th International AAAI Conference on Web and Social Media (ICWSM) (2020)
Abstract
Most individuals consider their friends to be more positive than themselves, exhibiting a sentiment paradox. Psychological research attributes this paradox to human cognition bias. With the goal to understand this phenomenon, we study sentiment paradoxes in social networks. Our work shows that social connections (friends, followees, or followers) of users are indeed generally (not illusively) more positive than the users themselves. Five existing sentiment paradoxes are identified at different network levels ranging from triads to large-scale communities. Empirical and theoretical evidence are provided to verify the observed and expected existence of such sentiment paradoxes. By investigating the relationships between the sentiment paradox and other well-developed network paradoxes, i.e., friendship paradox and activity paradox, we found that user sentiments are positively correlated to their number of social connections while hardly to their social activity. Finally, we demonstrate how the validated sentiment paradoxes can be used in turn to predict user sentiments.
SGCN: A Graph Sparsifier based on Graph Convolutional Networks
Conference PaperProceedings of the 24th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD) (2020)
Abstract
Graphs are ubiquitous across the globe and within science and engineering. With graphs growing in size, node classification on large graphs can be space and time consuming, even with powerful classifiers such as Graph Convolutional Networks (GCNs). Hence, some questions are raised, particularly, whether one can keep only some of the edges of a graph while maintaining prediction performance for node classification, or train classifiers on specific subgraphs instead of a whole graph with limited performance loss in node classification. To address these questions, we propose Sparsified Graph Convolutional Network (SGCN), a neural network graph sparsifier that sparsifies a graph by pruning some edges. We formulate sparsification as an optimization problem, which we solve by an Alternating Direction Method of Multipliers (ADMM) based solution. We show that sparsified graphs provided by SGCN can be used as inputs to GCN, leading to better or comparable node classification performance with that of original graphs in GCN, DeepWalk, and GraphSAGE. We provide insights on why SGCN performs well by analyzing its performance from the view of a low-pass filter.
The Spectral Zoo of Networks: Embedding and Visualizing Networks with Spectral Moments
Conference PaperProceedings of the 26th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) (2020)
Abstract
Network embedding techniques have been widely and successfully used in network-based applications such as node classification and link prediction. However, an ideal representation of a network should be both informative for prediction and be easy to interpret by users. In this paper, we introduce a spectral embedding method for a network, its Spectral Point, which is basically the truncated spectral moments of a network. We mathematically prove that spectral moments have close relationship with network structure (e.g. number of triangles and squares) and various network properties (e.g. degree distribution, clustering coefficient and network connectivity). Using spectral points, we introduce a visualizable and bounded 3D embedding space, where user can characterize different networks such as special graphs (e.g., cycles), or real-world networks from different categories (e.g., social or biological networks). We demonstrate that spectral points can be used for network identification (i.e., what network is this subgraph sampled from?) and the truncated spectral moments do not lose much predictive power.
ReCOVery: A Multimodal Repository for COVID-19 News Credibility Research
Conference PaperProceedings of the 29th ACM International Conference on Information and Knowledge Management (CIKM) (2020)
Abstract
First identified in Wuhan, China, in December 2019, the outbreak of COVID-19 has been declared as a global emergency in January, and a pandemic in March 2020 by the World Health Organization (WHO). Along with this pandemic, we are also experiencing an "infodemic" of information with low credibility such as fake news and conspiracies. In this work, we present ReCOVery, a repository designed and constructed to facilitate the studies of combating such information regarding COVID-19. We first broadly search and investigate ~2,000 news publishers, from which 61 are identified with extreme [high or low] levels of credibility. By inheriting the credibility of the media on which they were published, a total of 2,029 news articles on coronavirus, published from January to May 2020, are collected in the repository, along with 140,820 tweets that reveal how these news articles are spread on the social network. The repository provides multimodal information of news articles on coronavirus, including textual, visual, temporal, and network information. The way that news credibility is obtained allows a trade-off between dataset scalability and label accuracy. Extensive experiments are conducted to present data statistics and distributions, as well as to provide baseline performances for predicting news credibility so that future methods can be directly compared. Our repository is available at https://coronavirus-fakenews.com, which will be timely updated.
Exploiting Common Neighbor Graph for Link Prediction
Conference PaperProceedings of the 29th ACM International Conference on Information and Knowledge Management (CIKM) (2020)
Abstract
Link prediction aims to predict whether two nodes in a network are likely to get connected. Motivated by its applications, e.g., in friend or product recommendation, link prediction has been extensively studied over the years. Most link prediction methods are designed based on specific assumptions that may or may not hold in different networks, leading to link prediction methods that are not generalizable. Here, for the first time, we address this problem by proposing general link prediction methods that can capture network-specific patterns. Most link prediction methods rely on computing similarities between between nodes. By learning a $\gamma$-decaying model, the proposed methods can measure the pairwise similarities between nodes more accurately, even when only using common neighbor information, which is often used by current techniques.
Noise-Enhanced Community Detection
Conference PaperProceedings of the 31st ACM Conference on Hypertext and Social Media (HT) (2020)
Abstract
Community structure plays a significant role in uncovering the structure of a network. While many community detection algorithms have been introduced, improving the quality of detected communities is still an open problem. In many areas of science, adding noise improves system performance and algorithm efficiency, motivating us to also explore the possibility of adding noise to improve community detection algorithms. We propose a noise-enhanced community detection framework that improves communities detected by existing community detection methods. The framework introduces three noise methods to help detect communities better. Theoretical justification and extensive experiments on synthetic and real-world datasets show that our framework helps community detection methods find better communities.
Fake News Detection: An Interdisciplinary Research
Conference PaperCompanion Proceedings of The Web Conference 2019 (WWW) (2019)
Abstract
The explosive growth of fake news and its erosion to democracy, journalism and economy has increased the demand for fake news detection. To achieve efficient and explainable fake news detection, an interdisciplinary approach is required, relying on scientific contributions from various disciplines, e.g., social sciences, engineering, among others. Here, we illustrate how such multidisciplinary contributions can help detect fake news by improving feature engineering, or by providing well-justified machine learning models. We demonstrate how news content, news propagation patterns, and users' engagements with news can help detect fake news.
Fake News: Fundamental Theories, Detection Strategies and Challenges
TutorialProceedings of the 12th ACM International Conference on Web Search and Data Mining (WSDM) (2019)
Abstract
The explosive growth of fake news and its erosion to democracy, justice, and public trust increased the demand for fake news detection. As an interdisciplinary topic, the study of fake news encourages a concerted effort of experts in computer and information science, political science, journalism, social science, psychology, and economics. A comprehensive framework to systematically understand and detect fake news is necessary to attract and unite researchers in related areas to conduct research on fake news. This tutorial aims to clearly present (1) fake news research, its challenges, and research directions; (2) a comparison between fake news and other related concepts (e.g., rumors); (3) the fundamental theories developed across various disciplines that facilitate interdisciplinary research; (4) various detection strategies unified under a comprehensive framework for fake news detection; and (5) the state-of-the-art datasets, patterns, and models. We present fake news detection from various perspectives, which involve news content and information in social networks, and broadly adopt techniques in data mining, machine learning, natural language processing, information retrieval and social search. Facing the upcoming 2020 U.S. presidential election, challenges for automatic, effective and efficient fake news detection are also clarified in this tutorial.
The Impact of Graph Structure on Small-World Shortest Paths
Conference PaperProceedings of the 2019 International Conference on Social Computing, Behavioral-Cultural Modeling \& Prediction and Behavior Representation in Modeling and Simulation (SBP-BRiMS) (2019)
Abstract
Numerous studies have been devoted to modeling and estimating shortest-paths in complex networks. To maintain generality, these studies have neglected a common property of complex social networks; small-world phenomenon (colloquially stated as six degrees of separation). Based on the intuition behind the flow of information in smallworlds, we propose a small-world representation for social networks. In this new representation, we study the influence of different network measures on the shortest-paths. We perform a comprehensive analysis on a large set of local and global network measures and report our findings for various social networks. The results of our analyses show that: (1) shortest path lengths in small worlds are strongly correlated to the maximum degree centrality and the diameter. In fact, using these two features one can predict the average path length more accurately than using any other feature alone; (2) when nodes are ranked according to their average shortest-path lengths, we can approximate this ranking by a shifted standard normal distribution with minimum information loss. The shift can be estimated by the rank of the node with maximum local clustering coefficient, which can be computed in linear or constant time
Fake News Research: Theories, Detection Strategies, and Open Problems
TutorialProceedings of the 25th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) (2019)
Abstract
The explosive growth of fake news and its erosion to democracy, justice, and public trust increased the demand for fake news detection. As an interdisciplinary topic, the study of fake news encourages a concerted effort of experts in computer and information science, political science, journalism, social science, psychology, and economics. A comprehensive framework to systematically understand and detect fake news is necessary to attract and unite researchers in related areas to conduct research on fake news. This tutorial aims to clearly present (1) fake news research, its challenges, and research directions; (2) a comparison between fake news and other related concepts (e.g., rumors); (3) the fundamental theories developed across various disciplines that facilitate interdisciplinary research; (4) various detection strategies unified under a comprehensive framework for fake news detection; and (5) the state-of-the-art datasets, patterns, and models. We present fake news detection from various perspectives, which involve news content and information in social networks, and broadly adopt techniques in data mining, machine learning, natural language processing, information retrieval and social search. Facing the upcoming 2020 U.S. presidential election, challenges for automatic, effective and efficient fake news detection are also clarified in this tutorial.
Network Identification and Authentication
Conference PaperProceedings of the 2019 IEEE International Conference on Data Mining (ICDM) (2019)
Abstract
Research on networks is commonly performed using anonymized network data for various reasons such as protecting data privacy. Under such circumstances, it is difficult to verify the source of network data, which leads to questions such as: Given an anonymized graph, can we identify the network from which it is collected? Or if one claims the graph is sampled from a certain network, can we verify it? The intuitive approach is to check for subgraph isomorphism. However, subgraph isomorphism is NP-complete; hence, infeasible for most large networks. Inspired by biometrics studies, we address these challenges by formulating two new problems: network identification and network authentication. To tackle these problems, similar to research on human fingerprints, we introduce two versions of a network identity: (1) embedding-based identity and (2) distribution-based identity. We demonstrate the effectiveness of these network identities on various real-world networks. Using these identities, we propose two approaches for network identification. One method uses supervised learning and can achieve an identification accuracy rate of 94.7%, and the other, which is easier to implement, relies on distances between identities and achieves an accuracy rate of 85.5%. For network authentication, we propose two methods to build a network authentication system. The first is a supervised learner and provides a low false accept rate and the other method allows one to control the false reject rate with a reasonable false accept rate across networks. Our study can help identify or verify the source of network data, validate network-based research, and be used for network-based biometrics.
Sentiment Prediction in Social Networks
Conference PaperProceedings of the 2018 IEEE International Conference on Data Mining (ICDM) (2018)
Abstract
Sentiment analysis research has focused on using text for predicting sentiments without considering the unavoidable peer influence on user emotions and opinions. The lack of large-scale ground-truth data on sentiments of users in social networks has limited research on how predictable sentiments are from social ties. In this paper, using a large-scale dataset on human sentiments, we study sentiment prediction within social networks. We demonstrate that sentiments are predictable using structural properties of social networks alone. With social science and psychology literature, we provide evidence on sentiments being connected to social relationships at four different network levels, starting from the ego-network level and moving up to the whole-network level. We discuss emotional signals that can be captured at each level of social relationships and investigate the importance of structural features on each network levels. We demonstrate that sentiment prediction that solely relies on social network structure can be as (or more) accurate than text-based techniques. For the situations where complete posts and friendship information are difficult to get, we analyze the trade-off between the sentiment prediction performance and the available information. When computational resources are limited, we show that using only four network properties, one can predict sentiments with competitive accuracy. Our findings can be used to (1) validate the peer influence on user sentiments, (2) improve classical text-based sentiment prediction methods, (3) enhance friend recommendation by utilizing sentiments, and (4) help identify personality traits.
Representing Networks with 3D Shapes
Conference PaperProceedings of the 2018 IEEE International Conference on Data Mining (ICDM) (2018)
Abstract
There has been a surge of interest in machine learning in graphs, as graphs and networks are ubiquitous across the globe and within science and engineering: road networks, power grids, protein-protein interaction networks, scientific collaboration networks, social networks, to name a few. Recent machine learning research has focused on efficient and effective ways to represent graph structure. Existing graph representation methods such as network embedding techniques learn to map a node (or a graph) to a vector in a low-dimensional vector space. However, the mapped values are often difficult to interpret, lacking information on the structure of the network or its subgraphs. Instead of using a low-dimensional vector to represent a graph, we propose to represent a network with a 3-dimensional shape: the network shape. We introduce the first network shape, a Kronecker hull, which represents a network as a 3D convex polyhedron using stochastic Kronecker graphs. We present a linear time algorithm to build Kronecker hulls. Network shapes provide a compact representation of networks that is easy to visualize and interpret. They captures various properties of not only the network, but also its subgraphs. For instance, they can provide the distribution of subgraphs within a network, e.g., what proportion of subgraphs are structurally similar to the whole network? Using experiments on real-world networks, we show how network shapes can be used in various applications, from computing similarity between two graphs (using the overlap between network shapes of two networks) to graph compression, where a graph with millions of nodes can be represented with a convex hull with less than 40 boundary points.
User Identity Linkage across Online Social Networks: A Review.
Journal PaperACM SIGKDD Explorations Newsletter (Mar 2017)
Abstract
The increasing popularity and diversity of social media sites has encouraged more and more people to participate on multiple online social networks to enjoy their services. Each user may create a user identity, which can includes profile, content, or network information, to represent his or her unique public figure in every social network. Thus, a fundamental question arises -- can we link user identities across online social networks? User identity linkage across online social networks is an emerging task in social media and has attracted increasing attention in recent years. Advancements in user identity linkage could potentially impact various domains such as recommendation and link prediction. Due to the unique characteristics of social network data, this problem faces tremendous challenges. To tackle these challenges, recent approaches generally consist of (1) extracting features and (2) constructing predictive models from a variety of perspectives. In this paper, we review key achievements of user identity linkage across online social networks including stateof- the-art algorithms, evaluation metrics, and representative datasets. We also discuss related research areas, open problems, and future research directions for user identity linkage across online social networks.
Emotions in Social Networks: Distributions, Patterns, and Models
Conference PaperProceedings of the 2017 ACM Conference on Information and Knowledge Management (CIKM) (2017)
Abstract
Understanding the role emotions play in social interactions has been a central research question in the social sciences. However, the challenge of obtaining large-scale data on human emotions has left the most fundamental questions on emotions less explored: How do emotions vary across individuals, evolve over time, and are connected to social ties? We address these questions using a large-scale dataset of users that contains both their emotions and social ties. Using this dataset, we identify patterns of human emotions on five different network levels, starting from the user-level and moving up to the whole-network level. At the user-level, we identify how human emotions are distributed and vary over time. At the ego-network level, we find that assortativity is only observed with respect to positive moods. This observation allows us to introduce emotional balance, the "dual'' of structural balance theory. We show that emotional balance has a natural connection to structural balance theory. At the community-level, we find that community members are emotionally-similar and that this similarity is stronger in smaller communities. Structural properties of communities, such as their sparseness or isolatedness, are also connected to the emotions of their members. At the whole-network level, we show that there is a tight connection between the global structure of a network and the emotions of its members. As a result, we demonstrate how one can accurately predict the proportion of positive/negative users within a network by only looking at the network structure. Based on our observations, we propose the Emotional-Tie model -- a network model that can simulate the formation of friendships based on emotions. This model generates graphs that exhibit both patterns of human emotions identified in this work and those observed in real-world social networks, such as having a high clustering coefficient. Our findings can help better understand the interplay between emotions and social ties.
Users Joining Multiple Sites: Friendship and Popularity Variations across Sites
Journal PaperInformation Fusion (Mar 2016)
Abstract
Our social media experience is no longer limited to a single site. We use different social media sites for different purposes and our information on each site is often partial. By collecting complementary information for the same individual across sites, one can better profile users. These profiles can help improve online services such as advertising or recommendation across sites. To combine complementary information across sites, it is critical to understand how information for the same individual varies across sites. In this study, we aim to understand how two fundamental properties of users vary across social media sites. First, we study how user friendship behavior varies across sites. Our findings show how friend distributions for individuals change as they join new sites. Next, we analyze how user popularity changes across sites as individuals join different sites. We evaluate our findings and demonstrate how our findings can be employed to predict how popular users are likely to be on new sites they join.
The good, the bad, and the ugly: uncovering novel research opportunities in social media mining
Journal PaperInternational Journal of Data Science and Analytics 1.3-4 (2016), pp. 137--143
Abstract
Big data is ubiquitous and can only become bigger, which challenges traditional data mining and machine learning methods. Social media is a new source of data that is significantly different from conventional ones. Social media data are mostly user-generated, and are big, linked, and heterogeneous. We present the good, the bad and the ugly associated with the multi-faceted social media data and exemplify the importance of some original problems with real-world examples. We discuss bias in social media data, evaluation dilemma, data reduction, inferring invisible information, and big-data paradox. We illuminate new opportunities of developing novel algorithms and tools for data science. In our endeavor of employing the good to tame the bad with the help of the ugly, we deepen the understanding of ever growing and continuously evolving data and create innovative solutions with interdisciplinary and collaborative research of data science.
Evaluation without Ground Truth in Social Media Research
Journal PaperCommunications of the ACM (Jun 2015)
Abstract
With the rise of social media, user generated content has become available at an unprecedented scale. These massive collections of user generated content can help better understand billions of individuals. This data also enables novel scientific research in social sciences, anthropology, psychology, and economics at scale. Scientific research demands reproducible and independently verifiable findings. In social media research, scientific findings can be in form of behavioral patterns such as <i>individuals commented on this Facebook post because of its quality.</i> To validate such patterns one can survey the individuals that exhibited the pattern to verify if the patterns truly captured their intentions. This type of validation is known as evaluation with ground truth in data mining. However, social media users are scattered all across the globe. With no face-to-face access to individuals on social media, is it even possible to perform evaluation for social media research? In other words, how can we verify that the user behavioral patterns found are the `true patterns' of these individuals?
Sarcasm Detection on Twitter: A Behavioral Modeling Approach
Conference PaperProceedings of the 8th ACM International Conference on Web Search and Data Mining (WSDM) (2015)
Abstract
Sarcasm is a nuanced form of language in which individuals state the opposite of what is implied. With this intentional ambiguity, sarcasm detection has always been a challenging task, even for humans. Current approaches to automatic sarcasm detection rely primarily on lexical and linguistic cues. This paper aims to address the difficult task of sarcasm detection on Twitter by leveraging behavioral traits intrinsic to users expressing sarcasm. We identify such traits using the user's past tweets. We employ theories from behavioral and psychological studies to construct a behavioral modeling framework tuned for detecting sarcasm. We evaluate our framework and demonstrate its efficiency in identifying sarcastic tweets.
User Identification across Social Media
Journal PaperACM Transactions on Knowledge Discovery from Data (Oct 2015)
Abstract
People use various social media sites for different purposes. The information on each site is often partial. When sources of complementary information are integrated, a better profile of a user can be built. This profile can help improve online services such as advertising across sites. To integrate these sources of information, it is necessary to identify individuals across social media sites. This paper aims to address the cross-media user identification problem. We provide evidence on the existence of a mapping among identities of individuals across social media sites, study the feasibility of finding this mapping, and illustrate and develop means for finding this mapping. Our studies show that effective approaches that exploit information redundancies due to users' unique behavioral patterns can be utilized to find such a mapping. This study paves the way for analysis and mining across social networking sites, and facilitates the creation of novel online services across sites. In particular, recommending friends and advertising across networks, analyzing information diffusion across sites, and studying specific user behavior such as user migration across sites in social media are one of the many areas that can benefit from the results of this study
Real-Time Crisis Mapping using Language Distribution
Conference PaperProceedings of the 2015 IEEE International Conference on Data Mining (ICDM) (2015)
Abstract
With the increase in GPS-enabled devices, social media sites, such as Twitter, are quickly becoming a prime outlet for timely geo-spatial data. Such data can be leveraged to aid in emergency response planning and recovery operations. Unfortunately, the information overload poses significant difficulty to the quick discovery and identification of emergency situation areas. The system tackles this challenge by providing real-time mapping of influence areas based on automatic analysis of the flow of discussion using language distributions. The workflow is then further enhanced through the addition of keyword surprise mapping which projects the general divergence map onto specific task-level keywords for precise and focused response.
10 Bits of Surprise: Detecting Malicious Users with Minimum Information
Conference PaperProceedings of the 24th ACM International Conference on Information and Knowledge Management (CIKM) (2015)
Abstract
Malicious users are a threat to many sites and defending against them demands innovative countermeasures. When malicious users join sites, they provide limited information about themselves. With this limited information, sites can find it difficult to distinguish between a malicious user and a normal user. In this study, we develop a methodology that identifies malicious users with limited information. As information provided by malicious users can vary, the proposed methodology utilizes minimum information to identify malicious users. It is shown that as little as 10 bits of information can help greatly in this challenging task. The experiments results verify that this methodology is effective in identifying malicious users in the realistic scenario of limited information availability.
Am I More Similar to My Followers or Followees? Analyzing Homophily Effect in Directed Social Networks
Conference PaperProceedings of the 25th ACM Conference on Hypertext and Social Media (HT) (2014)
Abstract
Homophily is the theory behind the formation of social ties between individuals with similar characteristics or interests. Based on homophily, in a social network it is expected to observe a higher degree of homogeneity among connected than disconnected people. Many researchers use this simple yet effective principal to infer users' missing information and interests based on the information provided by their neighbors. In a directed social network, the neighbors can be further divided into followers and followees. In this work, we investigate the homophily effect in a directed network. To explore the homophily effect in a directed network, we study if a user's personal preferences can be inferred from those of users connected to her (followers or followees). We also study the effectiveness of each of these two groups on prediction one's preferences.
Finding Friends on a New Site Using Minimum Information
Conference PaperProceedings of the 2014 SIAM International Conference on Data Mining (SDM) (2014)
Abstract
With the emergence of numerous social media sites, individuals, with their limited time, often face a dilemma of choosing a few sites over others. Users prefer more engaging sites, where they can find familiar faces such as friends, relatives, or colleagues. Link prediction methods help find friends using link or content information. Unfortunately, whenever users join any site, they have no friends or any content generated. In this case, sites have no chance other than recommending random influential users to individuals hoping that users by befriending them create sufficient information for link prediction techniques to recommend meaningful friends. In this study, by considering social forces that form friendships, namely, influence, homophily, and confounding, and by employing minimum information available for users, we demonstrate how one can significantly improve random predictions without link or content information. In addition, contrary to the common belief that similarity between individuals is the essence of forming friendships, we show that it is the similarity that one exhibits to the friends of another individual that plays a more decisive role in predicting their future friendship.
Users Joining Multiple Sites: Distributions and Patterns
Conference PaperProceedings of the 8th International AAAI Conference on Web and Social Media (ICWSM) (2014)
Abstract
The rise of social media has led to an explosion in the number of possible sites users can join. However, this same profusion of social media sites has made it nearly impossible for users to actively engage in all of them simultaneously. Accordingly, users must make choices about which sites to use or to neglect. In this paper, we study users that have joined multiple sites. We study how individuals are distributed across sites, the way they select sites to join, and behavioral patterns they exhibit while selecting sites. Our study demonstrates that while users have a tendency to join the most popular or trendiest sites, this does not fully explain users' selections. We demonstrate that peer pressure also influences the decisions users make about joining emerging sites.
Abstract
With the rise of social media, information sharing has been democratized. As a result, users are given opportunities to exhibit different behaviors such as sharing, posting, liking, commenting, and befriending conveniently and on a daily basis. By analyzing behaviors observed on social media, we can categorize these behaviors into individual and collective behavior. Individual behavior is exhibited by a single user, whereas collective behavior is observed when a group of users behave together. For instance, users using the same hashtag on Twitter or migrating to another social media site are examples of collective behavior. User activities on social media generate behavioral data, which is massive, expansive, and indicative of user preferences, interests, opinions, and relationships. This behavioral data provides a new lens through which we can observe and analyze individual and collective behaviors of users.
Abstract
The growth of social media over the last decade has revolutionized the way individuals interact and industries conduct business. Individuals produce data at an unprecedented rate by interacting, sharing, and consuming content through social media. Understanding and processing this new type of data to glean actionable patterns presents challenges and opportunities for interdisciplinary research, novel algorithms, and tool development. Social Media Mining integrates social media, social network analysis, and data mining to provide a convenient and coherent platform for students, practitioners, researchers, and project managers to understand the basics and potentials of social media mining. It introduces the unique problems arising from social media data and presents fundamental concepts, emerging issues, and effective algorithms for network analysis and data mining. Suitable for use in advanced undergraduate and beginning graduate courses as well as professional short courses, the text contains exercises of different degrees of difficulty that improve understanding and help apply concepts, principles, and methods in various scenarios of social media mining.
Social Media Mining: Fundamental Issues and Challenges
TutorialProceedings of the 2013 IEEE International Conference on Data Mining (ICDM) (2013)
Abstract
Social media generates massive amounts of user-generated-content data. Such data differs from classic data and poses new challenges to data mining. This tutorial presents fundamental issues of social media mining, ranging from network representation to influence/diffusion modeling, elaborate state-of-the-art approaches of processing and analyzing social media data, and show how to utilize patterns to real-world applications, such as recommendation and behavior analytics. The tutorials designed for researchers, students and scholars interested in studying social media and social networks. No prerequisite is required for ICDM participants to attend this tutorial.
Connecting Users across Social Media Sites: A Behavioral-Modeling Approach
Conference PaperProceedings of the 19th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD) (2013)
Abstract
People use various social media for different purposes. The information on an individual site is often incomplete. When sources of complementary information are integrated, a better profile of a user can be built to improve online services such as verifying online information. To integrate these sources of information, it is necessary to identify individuals across social media sites. This paper aims to address the cross-media user identification problem. We introduce a methodology (MOBIUS) for finding a mapping among identities of individuals across social media sites. It consists of three key components: the first component identifies users' unique behavioral patterns that lead to information redundancies across sites; the second component constructs features that exploit information redundancies due to these behavioral patterns; and the third component employs machine learning for effective user identification. We formally define the cross-media user identification problem and show that MOBIUS is effective in identifying users across social media sites. This study paves the way for analysis and mining across social media sites, and facilitates the creation of novel online services across sites.
Whom Should I Follow? Identifying Relevant Users in Crisis
Conference PaperProceedings of the 24th ACM Conference on Hypertext and Social Media (HT) (2013)
Abstract
Social media is gaining popularity as a medium of communication before, during, and after crises. In several recent disasters, it has become evident that social media sites like Twitter and Facebook are an important source of information, and in cases they have even assisted in relief efforts. We propose a novel approach to identify a subset of active users during a crisis who can be tracked for fast access to information. Using a Twitter dataset that consists of 12.9 million tweets from 5 countries that are part of the "Arab Spring" movement, we show how instant information access can be achieved by user identification along two dimensions: user's location and the user's affinity towards topics of discussion. Through evaluations, we demonstrate that users selected by our approach generate more information and the quality of the information is better than that of users identified using state-of-the-art techniques.
Analyzing Behavior of the Influentials across Social Media
Book ChapterIn Behavior Computing: Modeling, Analysis, Mining and Decision, Springer London, 2012
Abstract
Studying these behavior patterns of users across different social media sites have many applications. If an individual exhibits same behavior on various social media sites then it can help predict his behavior on other social media websites by studying one social media site. Various social media sites can be clustered based on the behavior patterns of individuals. These clusters can help discover helpful and valuable trends. The activity of individuals can also help in explaining which social media sites are likely to get more activity for various groups of people. These patterns can be also used to explore marketing opportunities, study the movement of individuals on social media sites to focus on niche sites for unique opportunities.
Maximizing Benefits from Crowdsourced Data
Journal PaperComputational and Mathematical Organization Theory 18.3 (2012), pp. 257--279
Abstract
Crowds of people can solve some problems faster than individuals or small groups. A crowd can also rapidly generate data about circumstances affecting the crowd itself. This crowdsourced data can be leveraged to benefit the crowd by providing information or solutions faster than traditional means. However, the crowdsourced data can hardly be used directly to yield usable information. Intelligently analyzing and processing crowdsourced information can help prepare data to maximize the usable information, thus returning the benefit to the crowd. This article highlights challenges and investigates opportunities associated with mining crowdsourced data to yield useful information, as well as details how crowdsource information and technologies can be used for response-coordination when needed, and finally suggests related areas for future research.
Understanding User Migration Patterns in Social Media
Conference PaperProceedings of the 25th AAAI Conference on Artificial Intelligence (AAAI) (2011)
Abstract
Incredible growth of the social web over the last decade has created a flurry of new social media sites for the users to choose from. Users with a lack of time and resources to engage on these sites must inevitably choose specific sites among them to remain active on. This choice more often than not leads to user migration, a well studied phenomenon in fields such as sociology and psychology. Users are the valuable assets for any site as they generate revenue and contribute to the growth of the site. Hence, It is essential for social media sites to establish and understand the reasons for user migration to prevent it. In this paper, we investigate whether people migrate, and if they do, how they migrate. We formalize site migration to help identify the migration between popular social media sites and determine clear patterns of migration between sites. This work suggests that it is feasible to study migration patterns. These patterns can help understand social media sites and gauge their popularity to improve business intelligence and revenue generation through the retention of users.
Convergence of Influential Bloggers for Topic Discovery in the Blogosphere
Book ChapterIn Advances in Social Computing, Springer, 2010
Abstract
In this paper, we propose a novel approach to automatically detect .hot. or important topics of discussion in the blogosphere. The proposed approach is based on analyzing the activity of influential bloggers to determine specific points in time when there is a convergence amongst the influential bloggers in terms of their topic of discussion. The tool BlogTrackers, is used to identify influential bloggers and the Normalized Google Distance is used to define the similarity amongst the topics of discussion of influential bloggers. The key advantage of the proposed approach is its ability to automatically detect events which are important in the blogger community.
Sentiment Propagation in Social Networks: A Case Study in LiveJournal
Book ChapterIn Advances in Social Computing, Springer, 2010
Abstract
Social networking websites have facilitated a new style of communication through blogs, instant messaging, and various other techniques. Through collaboration, millions of users participate in millions of discussions every day. However, it is still difficult to determine the extent to which such discussions affect the emotions of the participants. We surmise that emotionally-oriented discussions may affect a given user.s general emotional bent and be reflected in other discussions he or she may initiate or participate in. It is in this way that emotion (or sentiment) may propagate through a network. In this paper, we analyze sentiment propagation in social networks, review the importance and challenges of such a study, and provide methodologies for measuring this kind of propagation. A case study has been conducted on a large dataset gathered from the LiveJournal social network. Experimental results are promising in revealing some aspects of the sentiment propagation taking place in social networks.
Connecting Corresponding Identities across Communities
Conference PaperProceedings of the Third International AAAI Conference on Web and Social Media (ICWSM) (2009)
Abstract
One of the most interesting challenges in the area of social computing and social media analysis is the so-called community analysis. A well known barrier in cross-community (multiple website) analysis is the disconnectedness of these websites. In this paper, our aim is to provide evidence on the existence of a mapping among identities across multiple communities, providing a method for connecting these websites. Our studies have shown that simple, yet effective approaches, which leverage social media's collective patterns can be utilized to find such a mapping. The employed methods successfully reveal this mapping with 66% accuracy.