Cross-language collaborative storytelling plays a vital role in children’s language learning and cultural development, fostering both expressive ability and intercultural awareness. Yet, in practice, children’s participation is often shallow, and facilitating such sessions places heavy cognitive and organizational burdens on coordinators, who must coordinate language support, maintain children’s engagement, and navigate cultural differences. To address these challenges, we conducted a formative study with coordinators to identify their needs and pain points, which guided the design of SparkTales, an intelligent support system for cross-language collaborative storytelling. SparkTales leverages both individual and common characteristics of participating children to provide coordinators with story frameworks, diverse questions, and comprehension-oriented materials, aiming to reduce coordinators’ workload while enhancing children’s interactive engagement. Evaluation results show that SparkTales not only significantly increases coordinators’ efficiency and quality of guidance but also improves children’s participation, providing valuable insights for the design of future intelligent systems supporting cross-language collaboration.

With the growing ubiquity of recommender systems (RSs), malicious manipulation through shilling attacks, where fake user profiles are injected to alter system outputs, poses increasing threats to system integrity. Existing attack methods often rely on simplified heuristics, require internal RS data, and most overlook user reviews, limiting their stealthiness, realism, and potential impact. Recently, LLM-based user agents are gaining traction in the RS community for their capabilities to simulate human behaviors like rating and review generation. In this context, we propose AgentSA, a low-knowledge shilling attack framework that employs such agents to manipulate recommendations through adversarial yet human-like interactions. We design targeted mechanisms to guide profile construction, memory retrieval, and action generation (including reviews) to maximize manipulation impact while maintaining behavioral camouflage. We evaluate the impact of these agents on various types of RSs and demonstrate that AgentSA consistently outperforms existing low-knowledge attack methods in both effectiveness and stealth. Our findings uncover a concerning new class of threats enabled by LLM-based agents, underscoring the pressing need to bolster RS security against such emerging risks.

As recommender systems become increasingly important components in online communities, studying their impact on these communities becomes ever more crucial. Facing the high costs and ethical risks of real-world social experiments, researchers construct recommendation simulators to study the interactions between recommender systems and users. However, existing simulators face challenges in providing universal, accurate, and scalable interaction modeling for various types of online communities involving millions of contents and users with diverse action types. To address these challenges, we propose RECOSIM, a simulation framework capable of offering efficient recommendation interaction simulations across a wide range of scenarios. RECOSIM decomposes the user agent into five fundamental modules: Encode Model, Decode Model, Activity Model, Scoring Model, and Generation Model, allowing for accurate and extensible modeling of user behavior and interaction dynamics. The recommender system agent adheres to established industry architectures, implementing three stages and four fundamental strategies, thereby improving generalizability across various platforms and the computational efficiency of simulation. Utilizing two real-world datasets (Weibo and Zhihu), we validate the accuracy and stability of each component and the overall framework of RECOSIM, demonstrating the reliability of RECOSIM as a simulation environment. Subsequently, we delve into analyzing the impact of the four fundamental recommendation strategies on online communities, providing design inspirations for enhancing user engagement and community growth.

Weiqi Wang, Limeng Cui, Xin Liu, Sreyashi Nag, Wenju Xu, Chen Luo, Sheikh Muhammad Sarwar, Yang Li, Hansu Gu, Hui Liu, Changlong Yu, Jiaxin Bai, Yifan Gao, Haiyang Zhang, Qi He, Shuiwang Ji, Yangqiu Song. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Online news platforms have become the primary source of information consumption, with recommender systems serving as critical gateways that shape public discourse through their algorithmic power, necessitating rigorous evaluation methodologies. Traditional offline evaluation methods struggle with evolving user behavior and dynamic system adaptation, while online experiments are costly, time-consuming, and ethically challenging. To address these challenges, this paper introduces EvalAgent, a large language model agent system for simulating real-world online news recommender systems. EvalAgent employs Stable Memory (StM) to model users' exploration-exploitation dynamics, mitigating noise from irrelevant interactions by analyzing the distribution density of news articles within the short-term memory, and incrementally maintains the long-term memory to capture users' high-level preferences, thereby enabling a consistent and reliable simulation of sustained interactions. It further incorporates an Environment Interaction Framework (EIF) to enable seamless engagement with real-world recommender systems. This approach yields a precise, scalable, and ethically responsible evaluation framework for news recommender systems. Comprehensive experiments and user studies substantiate EvalAgent's efficacy, with publicly available code to support ongoing research in recommender system evaluation.

Recent recommender systems aim to provide not only accurate recommendations but also explanations that help users understand them better. However, most existing explainable recommendations only consider the importance of content in reviews, such as words or aspects, and ignore the ordering relationship among them. This oversight neglects crucial ordering dimensions in the human decision-making process, leading to suboptimal performance. Therefore, in this paper, we propose Aspect Order Tree-based (AOTree) explainable recommendation method, inspired by the Order Effects Theory from cognitive and decision psychology, in order to capture the dependency relationships among decisive factors. We first validate the theory in the recommendation scenario by analyzing the reviews of the users. Then, according to the theory, the proposed AOTree expands the construction of the decision tree to capture aspect orders in users’ decision-making processes, and use attention mechanisms to make predictions based on the aspect orders. Extensive experiments demonstrate our method's effectiveness on rating predictions, and our approach aligns more consistently with the user’s decision-making process by displaying explanations in a particular order, thereby enhancing interpretability.

Recommendation algorithms rely on user historical interactions to deliver personalized suggestions, which raises significant privacy concerns. Federated recommendation algorithms tackle this issue by combining local model training with server-side model aggregation, where most existing algorithms use a uniform weighted summation to aggregate item embeddings from different client models. This approach has three major limitations: 1) information loss during aggregation, 2) failure to retain personalized local features, and 3) incompatibility with parameter-free recommendation algorithms. To address these limitations, we first review the development of recommendation algorithms and recognize that their core function is to share collaborative information, specifically the global relationship between users and items. With this understanding, we propose a novel aggregation paradigm named collaborative information aggregation, which focuses on sharing collaborative information rather than item parameters. Based on this new paradigm, we introduce the federated collaborative information aggregation (FedCIA) method for privacy-preserving recommendation. This method requires each client to upload item similarity matrices for aggregation, which allows clients to align their local models without constraining embeddings to a unified vector space. As a result, it mitigates information loss caused by direct summation, preserves the personalized embedding distributions of individual clients, and supports the aggregation of parameter-free models. Theoretical analysis and experimental results on real-world datasets demonstrate the superior performance of FedCIA compared with the state-of-the-art federated recommendation algorithms. Code is available at https://github.com/Mingzhe-Han/FedCIA.

Recommender systems leverage extensive user interaction data to model preferences; however, directly modeling these data may introduce biases that disproportionately favor popular items. In this paper, we demonstrate that popularity bias arises from the influence of propensity factors during training. Building on this insight, we propose a fair sampling (FS) method that ensures each user and each item has an equal likelihood of being selected as both positive and negative instances, thereby mitigating the influence of propensity factors. The proposed FS method does not require estimating propensity scores, thus avoiding the risk of failing to fully eliminate popularity bias caused by estimation inaccuracies. Comprehensive experiments demonstrate that the proposed FS method achieves state-of-the-art performance in both point-wise and pair-wise recommendation tasks. The code implementation is available at https://github.com/jhliu0807/Fair-Sampling.

LLM-based user agents, which simulate user interaction behavior, are emerging as a promising approach to enhancing recommender systems. In real-world scenarios, users' interactions often exhibit cross-domain characteristics and are influenced by others. However, the memory design in current methods causes user agents to introduce significant irrelevant information during decision-making in cross-domain scenarios and makes them unable to recognize the influence of other users' interactions, such as popularity factors. To tackle this issue, we propose a dual-layer memory architecture combined with a two-step fusion mechanism. This design avoids irrelevant information during decision-making while ensuring effective integration of cross-domain preferences. We also introduce the concepts of interest groups and group-shared memory to better capture the influence of popularity factors on users with similar interests. Comprehensive experiments validate the effectiveness of AgentCF++. Our code is available at https://github.com/jhliu0807/AgentCF-plus.

Due to the lack of explicit reasoning modeling, existing LLM-powered recommendations fail to leverage LLMs' reasoning capabilities effectively. In this paper, we propose a pipeline called CoT-Rec, which integrates two key Chain-of-Thought (CoT) processes-user preference analysis and item perception analysis-into LLM-powered recommendations, thereby enhancing the utilization of LLMs' reasoning abilities. CoT-Rec consists of two stages: (1) personalized information extraction, where user preferences and item perception are extracted, and (2) personalized information utilization, where this information is incorporated into the LLM-powered recommendation process. Experimental results demonstrate that CoT-Rec shows potential for improving LLM-powered recommendations. The implementation is publicly available at https://github.com/jhliu0807/CoT-Rec.

Sequential recommendation methods can capture dynamic user preferences from user historical interactions to achieve better performance. However, most existing methods only use past information extracted from user historical interactions to train the models, leading to the deviations of user preference modeling. Besides past information, future information is also available during training, which contains the ''oracle'' user preferences in the future and will be beneficial to model dynamic user preferences. Therefore, we propose an oracle-guided dynamic user preference modeling method for sequential recommendation (Oracle4Rec), which leverages future information to guide model training on past information, aiming to learn ''forward-looking'' models. Specifically, Oracle4Rec first extracts past and future information through two separate encoders, then learns a forward-looking model through an oracle-guiding module which minimizes the discrepancy between past and future information. We also tailor a two-phase model training strategy to make the guiding more effective. Extensive experiments demonstrate that Oracle4Rec is superior to state-of-the-art sequential methods. Further experiments show that Oracle4Rec can be leveraged as a generic module in other sequential recommendation methods to improve their performance with a considerable margin.

Personalized algorithms can inadvertently expose users to discomforting recommendations, potentially triggering negative consequences. The subjectivity of discomfort and the black-box nature of these algorithms make it challenging to effectively identify and filter such content. To address this, we first conducted a formative study to understand users' practices and expectations regarding discomforting recommendation filtering. Then, we designed a Large Language Model (LLM)-based tool named DiscomfortFilter, which constructs an editable preference profile for a user and helps the user express filtering needs through conversation to mask discomforting preferences within the profile. Based on the edited profile, DiscomfortFilter facilitates the discomforting recommendations filtering in a plug-and-play manner, maintaining flexibility and transparency. The constructed preference profile improves LLM reasoning and simplifies user alignment, enabling a 3.8B open-source LLM to rival top commercial models in an offline proxy task. A one-week user study with 24 participants demonstrated the effectiveness of DiscomfortFilter, while also highlighting its potential impact on platform recommendation outcomes. We conclude by discussing the ongoing challenges, highlighting its relevance to broader research, assessing stakeholder impact, and outlining future research directions.

Older adults (OAs) usually face various challenges when using smartphones due to their limited knowledge and the declines in memory and information processing capabilities. Many studies in HCI and CSCW communities have focused on supporting OAs to independently use smartphones. However, compared to independent exploration, support from younger family members (YFMs) has specific advantages in problem understanding, solution personalization, and security protection. However, OAs and YFMs generally have gaps in time, knowledge, and experience, affecting the efficiency of support and their experience. For this problem, we conduct a formative study to gather insights into OAs and YFMs' perspectives and expectations in the supporting procedure. Then we introduce chatbot to mediate the gaps between OAs and YFMs and build a system named WePilot to assist them to collaboratively solve smartphone usage problems. Evaluations with 12 pairs of participants (OA and corresponding YFM) suggest WePilot's strengths in improving problem solving efficiency and OAs and YFMs' experience. Based on these findings, we propose several insights into the future design of intergenerational technical support systems.

Although there have been automated approaches and tools supporting toxicity censorship for social posts, most of them focus on detection. Toxicity censorship is a complex process, wherein detection is just an initial task and a user can have further needs such as rationale understanding and content modification. For this problem, we conduct a need-finding study to investigate people's diverse needs in toxicity censorship and then build a ChatGPT-based censorship tool named DeMod accordingly. DeMod is equipped with the features of explainable De tection and personalized Mod ification, providing fine-grained detection results, detailed explanations, and personalized modification suggestions. We also implemented the tool and recruited 35 Weibo users for evaluation. The results suggest DeMod's multiple strengths like the richness of functionality, the accuracy of censorship, and ease of use. Based on the findings, we further propose several insights into the design of content censorship systems.

Performing user modeling on two or more social media platforms collaboratively and complementing each other (cross-site user modeling) has been a significant problem in the area of social media mining in recent years. The core of this problem is to get to know a person’s identities on multiple platforms and then train user models collaboratively among these platforms. However, for privacy protection, many people do not want their identities on different platforms to be linked and disclosed. For this problem, we set cross-site Content Preference Prediction as a task and propose a cross-site user modeling method without cross-site User Identity Linkage (UIL). The core thought borrowed from privacy-preserving recommender system research is to organize social media identities into groups to hide the identity linkage among platforms. Experiments on real-world datasets suggest that our method outperforms the existing cross-site user modeling methods with cross-site UIL regarding several metrics.

Temporal recommendation methods can achieve superior accuracy due to updating user/item embeddings continuously once obtaining new interactions. However, the randomness of user behaviors will introduce noises into the user interactions and cause the deviation in the modeling of user preference, resulting in sub-optimal performance. To this end, we propose NeuFilter, a robust temporal recommendation algorithm based on neural Kalman Filtering, to learn more accurate user and item embeddings with noisy interactions. Classic Kalman Filtering is time-consuming when applied to recommendation due to its covariance matrices. Thus, we propose a neural network solution to Kalman Filtering, so as to realize higher efficiency and stronger expressivity. Specifically, NeuFilter consists of three alternating units: 1) prediction unit, which predicts user and item embeddings based on their historical embeddings; 2) estimation unit, which updates user and item embeddings in a manner similar to Kalman Filtering; 3) correction unit, which corrects the updated user and item embeddings from estimation unit to ensure reliable estimation and accurate update. Experiments on two recommendation tasks show that NeuFilter can achieve higher accuracy compared with the state-of-the-art methods, while achieving high robustness. Moreover, our empirical studies on a node classification task further confirm the importance of handling noises in tasks on temporal graph, shedding a new light on temporal graph modeling.

Graph Signal Processing (GSP) has proven to be a highly effective and efficient tool for predicting user future interactions in recommender systems. However, current GSP methods recognize user interaction patterns based on the interactions of all users, so that the recognized interaction patterns are not fully user-matched and easily impacted by other users with different interaction behaviors, resulting in sub-optimal recommendation performance. To this end, we propose a hierarchical graph signal processing method (HiGSP) for collaborative filtering, which consists of two key modules: 1) the cluster-wise filter module that recognizes user unique interaction patterns merely from interactions of users with similar preferences, making the recognized patterns able to reflect user preference without being influenced by other users with different interaction behaviors, and 2) the globally-aware filter module that serves as a complementary to the cluster-wise filter module to recognize user general interaction patterns more effectively from all user interactions. By linearly combining these two modules, HiGSP can recognize user-matched interaction patterns, so as to model user preference and predict user future interactions more accurately. Extensive experiments on six real-world datasets demonstrate the superiority of HiGSP compared to other GCN-based and GSP-based recommendation methods in terms of efficacy and efficiency.

Jargon words are commonly used in the communication of online communities. These words are characterized by special and implicit meanings that can only be comprehended by a small group of users, which brings challenges to community regulation and user engagement. For this problem, we present JargonFM, a framework with multiple interpretation modes for jargon understanding in online communities. JargonFM is designed based on the scientific explanation framework and supports three interpretation modes: jargon category prediction based on a jargon classifier, similar word identification based on a jargon synonyms selector, and representative text selection based on an example sentence selector. A jargon interpreter was also implemented to demonstrate the usage and usefulness of the interpretation framework. Automatic and human evaluations suggest that JargonFM can explain jargon words more accurately and more efficiently than the existing interpretation methods, leading to its wide acceptance among the evaluation participants.

The rapid evolution of the web has led to an exponential growth in content. Recommender systems play a crucial role in human–computer interaction (HCI) by tailoring content based on individual preferences. Despite their importance, challenges persist in balancing recommendation accuracy with user satisfaction, addressing biases while preserving user privacy, and solving cold-start problems in cross-domain situations. This research argues that addressing these issues is not solely the recommender systems’ responsibility, and a human-centered approach is vital. We introduce the recommender system, assistant, and human (RAH) framework, an innovative solution with large language model (LLM)-based agents such as perceive, learn, act, critic, and reflect, emphasizing the alignment with user personalities. The framework utilizes the learn-act-critic loop and a reflection mechanism for improving user alignment. Using the real-world data, our experiments demonstrate the RAH framework's efficacy in various recommendation domains, from reducing human burden to mitigating biases and enhancing user control. Notably, our contributions provide a human-centered recommendation framework that partners effectively with various recommendation models.

Studying the evolution of online news communities is essential for improving the effectiveness of news recommender systems. Traditionally, this has been done through empirical research based on static data analysis. While this approach has yielded valuable insights for optimizing recommender system designs, it is limited by the lack of appropriate datasets and open platforms for controlled social experiments. This gap in the existing literature hinders a comprehensive understanding of the impact of recommender systems on the evolutionary process and its underlying mechanisms. As a result, suboptimal system designs may be developed that could negatively affect long-term utilities. In this work, we propose SimuLine, a simulation platform to dissect the evolution of news recommendation ecosystems and present a detailed analysis of the evolutionary process and underlying mechanisms. SimuLine first constructs a latent space well reflecting the human behaviors and then simulates the news recommendation ecosystem via agent-based modeling. Based on extensive simulation experiments and the comprehensive analysis framework consisting of quantitative metrics, visualization, and textual explanations, we analyze the characteristics of each evolutionary phase from the perspective of life-cycle theory and propose a relationship graph illustrating the key factors and affecting mechanisms. Furthermore, we explore the impacts of recommender system designing strategies, including the utilization of cold-start news, breaking news, and promotion, on the evolutionary process, which sheds new light on the design of recommender systems.

The emergence of online media has facilitated the dissemination of news, but has also introduced the problem of information overload. To address this issue, providing users with accurate and diverse news recommendations has become increasingly important. News possesses rich and heterogeneous content, and the factors that attract users to news reading are varied. Consequently, accurate news recommendation requires modeling of both the heterogeneous content of news and the heterogeneous user-news relationships. Furthermore, users’ news consumption is highly dynamic, which is reflected in the differences in topic concentration among different users and in the real-time changes in user interests. To this end, we propose a Heterogeneous Graph Neural Network with Personalized and Adaptive Diversity for News Recommendation (DivHGNN). DivHGNN first represents the heterogeneous content of news and the heterogeneous user-news relationships as an attributed heterogeneous graph. Then, through a heterogeneous node content adapter, it models the heterogeneous node attributes into aligned and fused node representations. With the proposed attributed heterogeneous graph neural network, DivHGNN integrates the heterogeneous relationships to enhance node representation for accurate news recommendations. We also discuss relation pruning, model deployment, and cold-start issues to further improve model efficiency. In terms of diversity, DivHGNN simultaneously models the variance of nodes through variational representation learning for providing personalized diversity. Additionally, a time-continuous exponentially decaying distribution cache is proposed to model the temporal dynamics of user real-time interests for providing adaptive diversity. Extensive experiments on real-world news datasets demonstrate the effectiveness of the proposed method.

Cross-site user identity linkage (UIL) aims to link the identities of the same person across different social media platforms. Social media practitioners and service providers can construct composite user portraits based on cross-site UIL, which helps understand user behavior holistically and conduct accurate recommendations and personalization. However, many social media users expect each profile to stay within the platform where it was created and thus do not want the identities of different platforms to be linked. For this problem, we first investigate the approaches people would like to use to defend against cross-site UIL and the corresponding challenges. Based on the findings, we build an adversarial framework, DeLink, based on the thoughts of adversarial text generation to help people improve their social media screen names to defend against cross-site UIL. DeLink can support both Chinese and English languages and has good generalizability to the varying numbers of social media accounts and different cross-site user identity linkage models. Extensive evaluations validate DeLink’s better performance, including a higher success rate, higher efficiency, less impact on human perception, and capability to defend against different cross-site UIL models.

Sequential recommendation demonstrates the capability to recommend items by modeling the sequential behavior of users. Traditional methods typically treat users as sequences of items, overlooking the collaborative relationships among them. Graph-based methods incorporate collaborative information by utilizing the user-item interaction graph. However, these methods sometimes face challenges in terms of time complexity and computational efficiency. To address these limitations, this paper presents AutoSeqRec, an incremental recommendation model specifically designed for sequential recommendation tasks. AutoSeqRec is based on autoencoders and consists of an encoder and three decoders within the autoencoder architecture. These components consider both the user-item interaction matrix and the rows and columns of the item transition matrix. The reconstruction of the user-item interaction matrix captures user long-term preferences through collaborative filtering. In addition, the rows and columns of the item transition matrix represent the item out-degree and in-degree hopping behavior, which allows for modeling the user's short-term interests. When making incremental recommendations, only the input matrices need to be updated, without the need to update parameters, which makes AutoSeqRec very efficient. Comprehensive evaluations demonstrate that AutoSeqRec outperforms existing methods in terms of accuracy, while showcasing its robustness and efficiency.

Click Through Rate (CTR) prediction plays an essential role in recommender systems and online advertising.It is crucial to effectively model feature interactions to improve the prediction performance of CTR models.However, existing methods face three significant challenges.First, while most methods can automatically capture high-order feature interactions, their performance tends to diminish as the order of feature interactions increases.Second, existing methods lack the ability to provide convincing interpretations of the prediction results, especially for high-order feature interactions, which limits the trustworthiness of their predictions.Third, many methods suffer from the presence of redundant parameters, particularly in the embedding layer.This paper proposes a novel method called Gated Deep Cross Network (GDCN) and a Field-level Dimension Optimization (FDO) approach to address these challenges.As the core structure of GDCN, Gated Cross Network (GCN) captures explicit high-order feature interactions and dynamically filters important interactions with an information gate in each order.Additionally, we use the FDO approach to learn condensed dimensions for each field based on their importance.Comprehensive experiments on five datasets demonstrate the effectiveness, superiority and interpretability of GDCN.Moreover, we verify the effectiveness of FDO in learning various dimensions and reducing model parameters.The code is available on https://github.com/anonctr/GDCN.

In dynamic interaction graphs, user-item interactions usually follow heterogeneous patterns, represented by different structural information, such as user-item co-occurrence, sequential information of user interactions and the transition probabilities of item pairs. However, the existing methods cannot simultaneously leverage all three structural information, resulting in suboptimal performance. To this end, we propose øurs, a triple structural information modeling method for accurate, explainable and interactive recommendation on dynamic interaction graphs. Specifically, øurs consists of 1) a dynamic ideal low-pass graph filter to dynamically mine co-occurrence information in user-item interactions, which is implemented by incremental singular value ecomposition (SVD); 2) a parameter-free attention module to capture sequential information of user interactions effectively and efficiently; and 3) an item transition matrix to store the transition probabilities of item pairs. Then, we fuse the predictions from the triple structural information sources to obtain the final recommendation results. By analyzing the relationship between the SVD-based and the recently emerging graph signal processing (GSP)-based collaborative filtering methods, we find that the essence of SVD is an ideal low-pass graph filter, so that the interest vector space in øurs can be extended to achieve explainable and interactive recommendation, making it possible for users to actively break through the information cocoons. Experiments on six public datasets demonstrated the effectiveness of øurs in accuracy, explainability and interactivity.

Many Click-Through Rate (CTR) prediction works focused on designing advanced architectures to model complex feature interactions but neglected the importance of feature representation learning, e.g., adopting a plain embedding layer for each feature, which results in sub-optimal feature representations and thus inferior CTR prediction performance. For instance, low frequency features, which account for the majority of features in many CTR tasks, are less considered in standard supervised learning settings, leading to sub-optimal feature representations. In this paper, we introduce self-supervised learning to produce high-quality feature representations directly and propose a model-agnostic Contrastive Learning for CTR (CL4CTR) framework consisting of three self-supervised learning signals to regularize the feature representation learning: contrastive loss, feature alignment, and field uniformity. The contrastive module first constructs positive feature pairs by data augmentation and then minimizes the distance between the representations of each positive feature pair by the contrastive loss. The feature alignment constraint forces the representations of features from the same field to be close, and the field uniformity constraint forces the representations of features from different fields to be distant. Extensive experiments verify that CL4CTR achieves the best performance on four datasets and has excellent effectiveness and compatibility with various representative baselines.

The collaborative filtering (CF) problem with only user-item interaction information can be solved by graph signal processing (GSP), which uses low-pass filters to smooth the observed interaction signals on the similarity graph to obtain the prediction signals. However, the interaction signal may not be sufficient to accurately characterize user interests and the low-pass filters may ignore the useful information contained in the high-frequency component of the observed signals, resulting in suboptimal accuracy. To this end, we propose a personalized graph signal processing (PGSP) method for collaborative filtering. Firstly, we design the personalized graph signal containing richer user information and construct an augmented similarity graph containing more graph topology information, to more effectively characterize user interests. Secondly, we devise a mixed-frequency graph filter to introduce useful information in the high-frequency components of the observed signals by combining an ideal low-pass filter that smooths signals globally and a linear low-pass filter that smooths signals locally. Finally, we combine the personalized graph signal, the augmented similarity graph and the mixed-frequency graph filter by proposing a pipeline consisting of three key steps: pre-processing, graph convolution and post-processing. Extensive experiments show that PGSP can achieve superior accuracy compared with state-of-the-art CF methods and, as a nonparametric method, PGSP has very high training efficiency.

Dynamic interaction graphs have been widely adopted to model the evolution of user-item interactions over time. There are two crucial factors when modelling user preferences for link prediction in dynamic interaction graphs: 1) collaborative relationship among users and 2) user personalized interaction patterns. Existing methods often implicitly consider these two factors together, which may lead to noisy user modelling when the two factors diverge. In addition, they usually require time-consuming parameter learning with back-propagation, which is prohibitive for real-time user preference modelling. To this end, this paper proposes FreeGEM, a parameter-free dynamic graph embedding method for link prediction. Firstly, to take advantage of the collaborative relationships, we propose an incremental graph embedding engine to obtain user/item embeddings, which is an Online-Monitor-Offline architecture consisting of an Online module to approximately embed users/items over time, a Monitor module to estimate the approximation error in real time and an Offline module to calibrate the user/item embeddings when the online approximation errors exceed a threshold. Meanwhile, we integrate attribute information into the model, which enables FreeGEM to better model users belonging to some under represented groups. Secondly, we design a personalized dynamic interaction pattern modeller, which combines dynamic time decay with attention mechanism to model user short-term interests. Experimental results on two link prediction tasks show that FreeGEM can outperform the state-of-the-art methods in accuracy while achieving over 36X improvement in efficiency. All code and datasets can be found in https://github.com/FudanCISL/FreeGEM.

CTR prediction has been widely used in the real world. Many methods model feature interaction to improve their performance. However, most methods only learn a fixed representation for each feature without considering the varying importance of each feature under different contexts, resulting in inferior performance. Recently, several methods tried to learn vector-level weights for feature representations to address the fixed representation issue. However, they only produce linear transformations to refine the fixed feature representations, which are still not flexible enough to capture the varying importance of each feature under different contexts. In this paper, we propose a novel module named Feature Refinement Network (FRNet), which learns context-aware feature representations at bit-level for each feature in different contexts. FRNet consists of two key components: 1) Information Extraction Unit (IEU), which captures contextual information and cross-feature relationships to guide context-aware feature refinement; and 2) Complementary Selection Gate (CSGate), which adaptively integrates the original and complementary feature representations learned in IEU with bit-level weights. Notably, FRNet is orthogonal to existing CTR methods and thus can be applied in many existing methods to boost their performance. Comprehensive experiments are conducted to verify the effectiveness, efficiency, and compatibility of FRNet.

To meet different social purposes, users usually share content related to the same topic or event to multiple social media platforms (cross-platform content sharing). As the differences of social norms and audiences among these social ecosystems, there are differences in the use of words and expressions in different platforms, resulting in different language styles among different platforms. In reality, it is usually difficult for users to grasp the consistency between the language style of posts to be published and that of a platform as the problem of context collapse. To address this problem, firstly, we conduct an study to investigate users' content sharing practices across two Chinese popular social media platforms (Douban and Weibo). The results indicate that: 1) there are significant linguistic differences between different platforms; 2) users' content sharing practices are personalized, and the style of their newly shared content is correlated with their historical posts. Secondly, based on the above findings, we propose a personalized cross-platform post style transfer model. The model can automatically transfer users' posts from one platform's language style to the target platform's language style, while preserving the content and reflecting users' personalized characteristics as much as possible. Experiments on the datasets collected from Douban and Weibo show that our model generally outperforms other comparison models on both style transfer and personalization metrics.

Auxiliary information, such as reviews, have been widely adopted to improve collaborative filtering (CF) algorithms, e.g., to boost the accuracy and provide explanations. However, most of the existing methods cannot distinguish between co-appearance and causality when learning from the reviews, so that they may rely on spurious correlations rather than causal relations in the recommendation — leading to poor generalization performance and unconvincing explanations. In this paper, we propose a Recommendation via Review Rationalization (R3) method including 1) a rationale generator to extract rationales from reviews to alleviate the effects of spurious correlations; 2) a rationale predictor to predict user ratings on items only from generated rationales; and 3) a correlation predictor upon both rationales and correlational features to ensure conditional independence between spurious correlations and rating predictions given causal rationales. Extensive experiments on real-world datasets show that the proposed method can achieve better generalization performance than state-of-the-art CF methods and provide causal-aware explanations even when the test data distribution changes.

Recommender systems are incremental in nature. Recent progresses in incremental recommendation rely on capturing the temporal dynamics of users/items from temporal interaction graphs, so that their user/item embeddings can evolve together with the graph structures. However, these methods are faced with two key challenges: 1) model training and/or updating are time-consuming and 2) new users/items cannot be effectively handled. To this end, we propose the fast incremental recommendation (FIRE) method from a graph signal processing perspective. FIRE is non-parametric which does not suffer from the time-consuming back-propagations as in previous learning-based methods, significantly improving the efficiency of model updating. In addition, we encode user/item temporal information and side information by designing new graph filters in FIRE, which can capture the temporal dynamics of users/items and address the cold-start issue for new users/items, respectively. Experimental studies on four popular datasets demonstrate that FIRE can improve the accuracy by a large margin and improve the model updating efficiency by at least 3X compared with the state-of-the-art incremental recommendation algorithms. The Code is available at https://github.com/Yaveng/FIRE.

Abstract The development of Information and Communication Technologies (ICT) and Web 2.0 promotes the emergence of diverse social ecosystems like social Internet of Things (IoT), social media and online communities. User-generated textual content (UGTC), which consists of unstructured texts, is the most important and common type of user-generated content in social ecosystems. UGTC in social ecosystems is generated according to two types of context information—global context (topics) and local context (semantic regularities). For UGTC modeling, topic models just consider global context but ignore semantic regularities, while semantic embedding models are on the opposite. So only utilizing topic models or semantic embedding models to model UGTC suffers from some drawbacks. For this problem, we propose a semantic embedding enhanced topic model named SEE-Twitter-LDA for accurately modeling UGTC in social ecosystems. The core of SEE-Twitter-LDA is that words are generated according to mutual semantic information of topics and semantic regularities. So global context and local context are jointly considered for UGTC modeling. By utilizing 553 098 tweets sampled from Twitter and 211 233 posts sampled from Weibo, we validate SEE-Twitter-LDA’s better performance on perplexity, topic divergence and topic coherence versus existing related models.

Social media users often suffer from the problem of content over-disclosure. Most existing studies attempt to solve this problem by recommending proper audiences for users when sharing content. However, the audience management strategy cannot filter out sensitive information from the post and narrow the scope of content permeation. On the contrary, this paper conducts research from the content perspective and aims to design a content censorship model to help users evaluate the publicity of a post and find the sensitive information from it. The user can revise the content accordingly to achieve goals of sensitive information protection and broader content permeation. For this intention, we first built a dataset to explore the factors related to the public level of a post and the sensitive information. Based on the findings, a novel personalized multi-task content censorship model was built using several state-of-the-art deep learning techniques such as Seq2Seq and Co-training. We also implemented a prototype, i.e. a Browser plugin-based content censorship tool, by utilizing Weibo as a research site. Our model and its prototype were evaluated through automatic and human evaluations. The automatic evaluation suggests that our model outperforms the baseline methods on several metrics including precision, recall, and F1-score. The human evaluation also reveals that our model and prototype play an important role in helping users identify sensitive information. Based on these results, we proposed several insights for the future design of the social media content censorship system.

Machine Translation (MT) has been a very useful tool to assist multilingual communication and collaboration. In recent years, by taking advantage of the exciting developments of neural networks and deep learning, the accuracy and speed of machine translation have been continuously improved. However, most machine translation methods and systems are data-driven. They tend to select a consensus response represented in training data, while a user's preferred linguistic style, which is important for translation comprehension and user experience, is ignored. For this problem, we aim to build a user-oriented personalized machine translation model in this paper. The model aims to learn each user's linguistic style from the textual content that is generated by her/him (User-Generated Textual Content, UGTC) in social media context and generate personalized translation results utilizing several state-of-the-art deep learning techniques like Transformer and pre-training. We also implemented a user-oriented personalized machine translator using Weibo as a case of the source of UGTC to provide a systematical implementation scheme of a user-oriented personalized machine translation system based on our model. The translator was evaluated by automatic evaluation in combination with human evaluation. The results suggest that our model can generate more personalized, natural and lively translation results and enhance the comprehensibility of translation results, which makes its generations more preferred by users versus general translation results.

User-generated contents (UGC) in social media are the direct expression of users’ interests, preferences, and opinions. User behavior prediction based on UGC has increasingly been investigated in recent years. Compared to learning a person’s behavioral patterns in each social media site separately, jointly predicting user behavior in multiple social media sites and complementing each other (cross-site user behavior prediction) can be more accurate. However, cross-site user behavior prediction based on UGC is a challenging task due to the difficulty of cross-site data sampling, the complexity of UGC modeling, and uncertainty of knowledge sharing among different sites. For these problems, we propose a Cross-Site Multi-Task (CSMT) learning method to jointly predict user behavior in multiple social media sites. CSMT mainly derives from the hierarchical attention network and multi-task learning. Using this method, the UGC in each social media site can obtain fine-grained representations in terms of words, topics, posts, hashtags, and time slices as well as the relevances among them, and prediction tasks in different social media sites can be jointly implemented and complement each other. By utilizing two cross-site datasets sampled from Weibo, Douban, Facebook, and Twitter, we validate our method’s superiority on several classification metrics compared with existing related methods.

Graph neural networks (GNN) recently achieved huge success in collaborative filtering (CF) due to the useful graph structure information. However, users will continuously interact with items, which causes the user-item interaction graphs to change over time and well-trained GNN models to be out-of-date soon. Naive solutions such as periodic retraining lose important temporal information and are computationally expensive. Recent works that leverage recurrent neural networks to keep GNN up-to-date may suffer from the "catastrophic forgetting'' issue, and experience a cold start with new users and items. To this end, we propose the incremental graph convolutional network (IGCN) --- a pure graph convolutional network (GCN) based method to update GNN models when new user-item interactions are available. IGCN consists of two main components: 1) a historical feature generation layer, which generates the initial user/item embedding via model agnostic meta-learning and ensures good initial states and fast model adaptation; 2) a temporal feature learning layer, which first aggregates the features from local neighborhood to update the embedding of each user/item within each subgraph via graph convolutional network and then fuses the user/item embeddings from last subgraph and current subgraph via incremental temporal convolutional network. Experimental studies on real-world datasets show that IGCN can outperform state-of-the-art CF algorithms in sequential recommendation tasks.

Many popular social network sites (SNSs) provide the post-syncing functionality, which allows users to synchronize posts automatically among different SNSs. Nowadays there exists divergence on this functionality from the view of sink SNS. The key to solving this problem is to understand the characteristics of users’ post-syncing practice and goals and evaluate whether they are consistent with an SNS’s norms, cultures, and goals. However, studying and understanding the characteristics of post-syncing practice and goal are challenging tasks as a result of the difficulty of data sampling and the complexity of post-syncing behavior. In this article, we focus on investigating this question by quantitative analysis in combination with qualitative analysis. In the quantitative study, by utilizing 211,233 synced-posts sampled from Weibo, we aim to investigate characteristics of post-syncing from three perspectives: user, content, and goal. The results suggest that post-syncing plays an important role in exhibiting one’s current activities, creations, and skills as well as advertisements but involves a risk of exhibiting personal sensitive profiles. To understand the results, we present an interview-based qualitative study based on thematic analysis. It indicates that the publicity, urgency, and remarkableness of contents and differences of social affordances and social circles between sink SNS and source SNS as well as the one-time consent of post-syncing authentication jointly account for the major role of post-syncing. Based on these results, we propose insights for post-syncing functionality’s adoption, design, and promotion.

Abstract Question‐Answering (Q&A) services provide internet users with platforms to exchange knowledge and ideas. The development of Q&A sites, or Community Question Answering (CQA), mainly depends on the high‐quality content continuously contributed by users with high‐level expertise, who can be recognized as experts. Expert finding is an important task for the authorities of Q&A communities to encourage commitment. In a highly competitive market environment, CQA managers have to take measures to retain and nurture users, especially superior contributors. However, current expertise scoring techniques adopted in CQA often give much credit to very active users and fail to identify real experts. This study aims to develop a robust and practical expert identification framework for Q&A communities, by combining well‐designed expertise scoring technique and probabilistic clustering model. With regard to expert identification, a numerical metric of users' expertise is developed as the optimal expert finding strategy, and a clustering algorithm based on Gaussian‐Gamma mixture model (GGMM) is proposed to efficiently distinguish experts from nonexperts. In the experiments, the proposed method is applied to real‐world datasets collected from subcommunities of Stack Exchange Q&A networks. Results obtained from comparative experiments show that our method achieves better performance than the state‐of‐the‐art methods and demonstrate the effectiveness of the proposed framework. The analysis shows that the framework which combines the proposed expertise scoring technique and Gaussian–Gamma mixture clustering model is capable of detecting excellent domain problem‐solving experts who exhibit both domain interest and expertise.

People tend to utilize multiple social network sites (SNSs) simultaneously to maintain some social relationships, which results that there are many overlapping relationships and interactions among SNSs. Although many studies have focused on social interaction and cross-SNS user footprint analysis and understanding, little research investigates social interaction from a perspective of two or more SNSs, and the interplay of interaction among SNSs has been unknown. In this paper, we aim to explore whether interaction building in a new SNS hinders the interaction frequency in an existing site, and if so, what kinds of users and relationships’ interactions are more or less likely to be affected. For these questions, we sampled 7,015 pairs of overlapping identities, 23,590 pairs of overlapping relationships and 6,771 pairs of overlapping interactions from Weibo and Douban and made analysis by combining multiple methods like Regression Discontinuity Design and Random-effects Negative Binomial Regression model. Our results suggest that no matter from the perspective of individuals or from the perspective of relationships, interaction construction in a new SNS is detrimental to interaction frequency in an existing site. Based on our findings, we also propose several valuable insights about how to enhance social interaction and promote its retention when users are involved into interacting practice in multiple platforms.

Online question and answer (Q&A) sites are platforms allowing users to seek information or social support, which is sometimes socially uncomfortable. A major challenge in Q&A sites is how to encourage user contribution. Theory of linguistic politeness informs us that politeness is able to ease the tension resulting from face-threatening acts (FTA), and it is applicable to Q&A sites. However, Stack Overflow (SO), one of the most popular Q&A site for programmers, denies the use of linguistic politeness (e.g., greetings, salutations, and gratitude expressions), aiming at building a nonsocial platform. This paper focuses on a specific politeness strategy-expressing gratitude-in Q&A sites, and examines its effect on answerers' future contributions from the aspects of quality. Results indicate that gratitude expressions in comments can motivate answerers to generate higher quality content.

Gradient-based learning methods such as stochastic gradient descent are widely used in matrix approximation-based collaborative filtering algorithms to train recommendation models based on observed user-item ratings. One major difficulty in existing gradient-based learning methods is determining proper learning rates, since model convergence would be inaccurate or very slow if the learning rate is too large or too small, respectively. This paper proposes AdaError, an adaptive learning rate method for matrix approximation-based collaborative filtering. AdaError eliminates the need of manually tuning the learning rates by adaptively adjusting the learning rates based on the noisiness level of user-item ratings, using smaller learning rates for noisy ratings so as to reduce their impact on the learned models. Our theoretical and empirical analysis shows that AdaError can improve the generalization performance of the learned models. Experimental studies on the MovieLens and Netflix datasets also demonstrate that AdaError outperforms state-of-the-art adaptive learning rate methods in matrix approximation-based collaborative filtering. Furthermore, by applying AdaError to the standard matrix approximation method, we can achieve statistically significant improvements over state-of-the-art collaborative filtering methods in both rating prediction accuracy and top-N recommendation accuracy.

Credit scoring plays an important role in financial institutions and debt based crowdfunding platforms as well as peer to peer lending platforms. In the last few years, adopting ensemble methods for credit scoring has become much more popular. However, the performance of ensemble methods is easily affected by the parameter settings and the number of base classifiers. Ensemble classification based on clustering is able to determine the best number of base classifiers automatically by clustering and find optimal parameter settings for base classifiers by training them individually on the training subsets combined by clusters. By this way, the adverse effect of manually setting the parameters and the number of base classifiers can be avoided. However, the different contributions of attributes to the distance metrics are not considered in conventional clustering methods, which may decrease the performance of ensemble classifiers based on them. Moreover, unbalanced training subsets decrease the performance of base classifiers, which results in the bad performance of ensemble classifiers. In our approach, to address the above problems, we first assign different weights to different variables when measuring the distance between two instances in the clustering step, and then adopt Subagging resampling method to deal with unbalanced training subsets in the training process. Experimental results show that our approach can improve the performance of the ensemble classifier.

In the past decade, online news consumption has been steadily growing. New articles are published every few minutes, and user preferences are also constantly changing. Traditional recommender systems update model at regular intervals, which cannot adjust recommendation list dynamically according to the changes of user preferences. In this paper, we propose a hybrid recommendation model which contains two key components: incremental update item-based collaborative filtering (CF) and latent semantic analysis based relative term frequency algorithms. The hybrid recommendation model adjusts recommendation list dynamically by updating similarity table of items incrementally in incremental update item-based CF module, moreover, combining collaborative filtering and content-based algorithm ensures the relevance of recommendation articles. Results show that our proposed hybrid recommendation model outperforms traditional recommender approaches.

Nowadays people have many overlapping relationships across different social network sites. Will the communication frequency of a relationship on a focal SNS increase or decrease after the corresponding parties begin interacting with each other on a new SNS? What kinds of relationships and parties on the focal SNS are more robust to relationship overlapping across different sites? We conducted a case study on two Chinese popular social network sites (Weibo and Douban). Our results indicate that relationships' interactions on a new SNS have negative effect on the parties' communication frequency on the focal SNS. The communication frequency of older relationships on the focal SNS is more susceptible to the influence of the interacting on the new SNS, while relationships with more common groups and more extensive posts are less likely to be influenced. Our findings imply opportunities for SNS designers to strengthen their existing fragile ties and help users to build much more robust relationships.

Smart environment has been evolved as a hot research topic with the development of machine learning algorithms and wireless communication technologies. Existing smart home solutions usually need inhabitants to operate device controllers directly. However, for people with disabilities, it is inconvenient and difficult to perform such manual operations. Therefore, it is important to develop automatic and intelligent services to reduce operation inconvenience and improve comfort level. In this paper, an Improved Hidden Markov Model (IHMM) is presented to support personalized behavior prediction for people with disabilities. The model can learn behavior patterns of users and provide services to inhabitants automatically. Moreover, by breaking the time invariant hypothesis in Hidden Markov Model, we incorporate time information as positions of states, and develop a temporal state transition matrix to replace the fixed state transition matrix to demonstrate the probabilities of state transitions. As a result, different values of daily temperature sections are characterized and identified as hidden variables, which guide user activities. Evaluation of proposed work has shown that IHMM improves the prediction accuracy by at least 10% compared to the traditional HMM.

As Sina Weibo is becoming one of the most popular online social networking platforms in China, there is growing need in extracting knowledge and analyzing user behavior out of it. Influential user discovery has been quickly identified as an important application for better marketing and advertising purposes. In order to analyze user influence, we construct a user influence actor-network based on Actor-Network Theory, in which both users and users' microblogs are seen as actors and their interactions are treated as networks. User influence is then evolved into network effects caused by both microblogs and interactions, which are characterized as the Quality Index and Popularity Index of microblog respectively that can be well evaluated through models proposed in this paper. Furthermore, by observing that one user may show distinctive influence in different domains, we build our model domain specifically. With full consideration to quality, popularity and domain categories of users' microblogs, we propose a novel domain-specific user influence measurement, ANTRank, in Sina Weibo based on Actor-Network Theory. Experimental results show that the ANTRank performs well with the user influence evaluation.

With the development of web 2.0, online community websites have become one of the most interactive ways for disabilities to communicate with others and understand the world. However, new disabled users may sometimes find it difficult to learn about existing communities and the topics they feel interested in are drastically different from normal users. On the other hand, influential users have already developed their impact in different topics within communities and are hence able to help disabilities quickly adapt to the communities. As the result, a method that can recommend influential users in diverse topics is necessary. However, traditional approaches to identify influential users are not very comprehensive and accurate without the consideration of topic diversity. In this paper, we present an evaluation index system and a SVD based influence model which can evaluate influence comprehensively. The discussions in online communities are then classified by the topics that attract disabilities most. Finally, a combined influential users recommendation with diverse topics is proposed. We conduct experiments on an online disability community dataset and results show that our approach significantly outperforms other traditional methods.

Online social networks (OSNs) such as Twitter provide a good platform for event discussions. Recent research [26][25] as shown that event discussions in OSNs are diverse and innovative and encourage public engagement in events. Although much research has been conducted in OSNs to track and detect events, there has been limited research on detecting or understanding the event context. Event context helps to better predict users' participation in events, identify relations among events, and recommend friends who share similar event context.

Location information is becoming prevalent in today's online social networks (OSNs), which raises special privacy concerns with regard to both location sharing and its applications. Even when no explicit location is disclosed by a user, it is possible to geolocate the user through his/her social context, e.g., status updates and social relationships in OSNs. To demonstrate this, we propose GeoFind, which accurately identifies users' geographic regions through effective fusion (re-ranking) of (1) text-based ranking using geo-sensitive textual features and (2) structure-based ranking using maximum likelihood estimation (MLE) of geotagged friends. Evaluation results using 0.8 million geotagged Twitter users over a 3-month period demonstrate that GeoFind outperforms state-of-the-art techniques, with significant reduction of estimation error (25% of average error, 66% of median error). The potential of improving location accuracy through the fusion of multiple data types calls for a re-examination of existing privacy protection policies and mechanisms.

Outline social media networks (OSMNs) such as Twitter provide great opportunities for public engagement and event information dissemination. Event-related discussions occur in real time and at the worldwide scale. However, these discussions are in the form of short, unstructured messages and dynamically woven into daily chats and status updates. Compared with traditional news articles, the rich and diverse user-generated content raises unique new challenges for tracking and analyzing events. Effective and efficient event modeling is thus essential for real-time information-intensive OSMNs. In this work, we propose ETree, an effective and efficient event modeling solution for social media network sites. Targeting the unique challenges of this problem, ETree consists of three key components: (1) an n-gram based content analysis technique for identifying core information blocks from a large number of short messages, (2) an incremental and hierarchical modeling technique for identifying and constructing event theme structures at different granularities, and (3) an enhanced temporal analysis technique for identifying inherent causalities between information blocks. Detailed evaluation using 3.5 million tweets over a 5-month period demonstrates that ETree can efficiently generate high-quality event structures and identify inherent causal relationships with high accuracy.

Interactive TV/arcade games have been entertaining people for over 50 years. Nowadays a large number of legacy TV/arcade games have been ported to new platforms such as PCs by emulation. However, they generally require that the players be co-located to interact with one computer that emulates the game. This paper proposes a novel approach to turning those single-computer games into multi-computer games such that multiple players can play their favorite legacy games in real time over a computer network. The main challenge in this work is how to synchronize multiple replicas of a game without semantic knowledge about or modifications to the game. We have developed a novel synchronization algorithm and a working system to validate the ideas. In this paper we present the approach, especially the synchronization algorithm, and evaluate its effectiveness under a variety of network conditions. In future research we will extend this work on mobile devices.