IEEE Transactions on Computational Social Systems最新文献_第2页

MHFS: Multimodal Haptic Feedback System for Enhanced Immersive Experience MHFS：增强沉浸式体验的多模态触觉反馈系统

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2026-01-01 DOI: 10.1109/TCSS.2025.3640775

Zihe Zhao;Hongbei Meng;Cong Li;Wenyu Yang;Shengbo Wang;Xuhang Chen;Chenyu Tang;Yu Pan;Arokia Nathan;Shuo Gao

{"title":"MHFS: Multimodal Haptic Feedback System for Enhanced Immersive Experience","authors":"Zihe Zhao;Hongbei Meng;Cong Li;Wenyu Yang;Shengbo Wang;Xuhang Chen;Chenyu Tang;Yu Pan;Arokia Nathan;Shuo Gao","doi":"10.1109/TCSS.2025.3640775","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3640775","url":null,"abstract":"Haptic feedback devices provide a sense of immersion that makes them play a key role in the domain of human-environment interactivities, such as in telemedicine, teleoperation, and online education. However, widely used unimodal or limited-modal approaches, such as vibrotactile, electrotactile, force-feedback, thermal, and emerging mid-air ultrasound/magnetorheological systems, are merely capable of simulating few virtual materials, exhibit a deficiency in fidelity, and necessitate substantial training expenditures. To address these limitations, we propose a multivirtual object recognition system based on multimodal haptic feedback and develop an evaluation and training scheme. The system is based on electrohaptic, vibrohaptic, and thermohaptic, acquires four haptic perception modalities, roughness, sharpness, hardness, and temperature, and simulates the haptic sensation of eight different materials. In experiments, the fidelity alignment of haptic feelings of real and virtual objects is 85.8%, and after averaged 2.8 trainings, the recognition accuracy reaches 100%. Our research not only improves the realism of haptic feedback but also significantly reduces the training cost, providing a more cost-effective and practical solution for user experience enhancement in virtual reality (VR) and meta-universe applications.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6656-6669"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097177","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Modeling Multi-Grained User Interests for Sequential Recommendation 为顺序推荐建模多粒度用户兴趣

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2026-03-05 DOI: 10.1109/TCSS.2026.3675684

Bin Wu;Xiaowen Yin;Xun Su;Mingliang Xu

{"title":"Modeling Multi-Grained User Interests for Sequential Recommendation","authors":"Bin Wu;Xiaowen Yin;Xun Su;Mingliang Xu","doi":"10.1109/TCSS.2026.3675684","DOIUrl":"https://doi.org/10.1109/TCSS.2026.3675684","url":null,"abstract":"Sequential recommendation has garnered considerable attention in recent years, which aims to infer the next desired item for each user based on their historical behavior data. Despite the remarkable progress made by previous methods, we believe that they still have two limitations: 1) they mainly focus on extracting pairwise transition pattern between two items within each individual sequence, how to fully capture the collective influence of a group of items is less explored; and 2) they pay insufficient attention to the intricate dependencies among items across different user behavior sequences, which may lead to suboptimal performance. In response to this, we propose a novel solution for sequential recommendation, termed the multi-grained user interest learning model (MGUI), which captures dynamic user interests by modeling item transition patterns at different levels. Specifically, our designed time interval-aware gating module aims to highlight the individual influence of each recent item on the target behavior, while the devised capsule-guided transformer module aims to capture the collective influence of a group of recent items on the next-to-visit item. These two modules explicitly exploit item transition patterns within a sequence at the individual- and group-level. Furthermore, we reorganize all users’ behavior sequences into a personalized item transition graph and devise a graph transition propagation module upon this graph to capture global-level dependencies among items across different sequences. Finally, we design a user-specific mixture-of-experts strategy to dynamically determine the importance of each module for each user. Extensive experiments across four recommendation scenarios demonstrate the superiority of MGUI over recent state-of-the-art recommendation models. In-depth analyses verify the effectiveness of each part in MGUI.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6145-6160"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097521","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

TimeResLSTM: A Hybrid Deep Learning Architecture for Short-Term Traffic Flow Forecasting TimeResLSTM：一种用于短期交通流量预测的混合深度学习架构

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2026-01-14 DOI: 10.1109/TCSS.2025.3642560

Vikas Chauhan;Aruna Tiwari

{"title":"TimeResLSTM: A Hybrid Deep Learning Architecture for Short-Term Traffic Flow Forecasting","authors":"Vikas Chauhan;Aruna Tiwari","doi":"10.1109/TCSS.2025.3642560","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3642560","url":null,"abstract":"Short-term traffic flow forecasting has emerged as an important aspect of the intelligent transportation system. This helps to resolve the traffic congestion problem and provides immediate route planning. Hybrid deep learning architectures have gained popularity by providing efficient solutions to traffic flow forecasting problems. In this article, we propose a hybrid deep-learning architecture TimeResLSTM that consists of the residual connections to learn the additional traffic patterns while resolving the exploding gradient problem. The proposed architecture has the capability to capture the periodic and nonperiodic patterns in the traffic data and it enhances the traffic flow forecasting performance. The proposed architecture consists of four modules TimeResLSTM, ConvLSTM, and two BiDirectional gated recurrent unit (Bi-GRU) modules to capture the periodic nonperiodic spatial, temporal, daily, and weekly patterns. The attention mechanism is also incorporated in the BiGRU modules to capture the intricate characteristics of the traffic flow. The fusion of the output of all four modules provides advantages of these modules and improves the forecasting results of short-term traffic flow forecasting. The TimeResLSTM architecture is evaluated on the real-world traffic dataset for the urban and freeway areas and outperforms the existing deep learning and hybrid deep learning architectures in terms of mean absolute error, mean absolute percentage error, and root means square error.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6258-6268"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097482","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Unpacking Generative AI in Education: Computational Modeling of Teacher and Student Perspectives in Social Media Discourse 在教育中打开生成人工智能的包装：社交媒体话语中教师和学生视角的计算建模

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2025-12-18 DOI: 10.1109/TCSS.2025.3630587

Paulina DeVito;Akhil Vallala;Sean Mcmahon;Yaroslav Hinda;Benjamin Thaw;Hanqi Zhuang;Hari Kalva

{"title":"Unpacking Generative AI in Education: Computational Modeling of Teacher and Student Perspectives in Social Media Discourse","authors":"Paulina DeVito;Akhil Vallala;Sean Mcmahon;Yaroslav Hinda;Benjamin Thaw;Hanqi Zhuang;Hari Kalva","doi":"10.1109/TCSS.2025.3630587","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3630587","url":null,"abstract":"Generative AI (GAI) technologies are quickly reshaping the educational landscape. As adoption accelerates, understanding how students and educators perceive these tools is essential. This study presents one of the most comprehensive analyses to date of stakeholder discourse dynamics on GAI in education using social media data. Our dataset includes 1199 Reddit posts and 13 959 corresponding top-level comments. We apply sentiment analysis, topic modeling, and author classification. To support this, we propose and validate a modular framework that leverages prompt-based large language models (LLMs) for analysis of online social discourse, and we evaluate this framework against classical natural language processing (NLP) models. Our GPT-4o pipeline consistently outperforms prior approaches across all tasks. For example, it achieved 90.6% accuracy in sentiment analysis against gold-standard human annotations. Topic extraction uncovered 12 latent topics in the public discourse with varying sentiment and author distributions. Teachers and students convey optimism about GAI’s potential for personalized learning and productivity in higher education. However, key differences emerged: students often voice distress over false accusations of cheating by AI detectors, while teachers generally express concern about job security, academic integrity, and institutional pressures to adopt GAI tools. These contrasting perspectives highlight the tension between innovation and oversight in GAI-enabled learning environments. Our findings suggest a need for clearer institutional policies, more transparent GAI integration practices, and support mechanisms for both educators and students. More broadly, this study demonstrates the potential of LLM-based frameworks for modeling stakeholder discourse within online communities.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6584-6596"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097505","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Federated Recommendation Model Integrating Social Relationships and Knowledge Graph 集成社会关系和知识图谱的联邦推荐模型

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2025-11-27 DOI: 10.1109/TCSS.2025.3630162

Junhua Gu;Lingyu Zhang;Cheng Liu;Suqi Zhang;Qiqi Liu

{"title":"Federated Recommendation Model Integrating Social Relationships and Knowledge Graph","authors":"Junhua Gu;Lingyu Zhang;Cheng Liu;Suqi Zhang;Qiqi Liu","doi":"10.1109/TCSS.2025.3630162","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3630162","url":null,"abstract":"Integrating social relationships and knowledge graph (KG) in centralized recommendations is an effective method to mitigate data sparsity and cold-start problems. However, current research has only attempted to include either social relationships or a KG in federated recommendation systems (FedRS), failing to utilize all available user and item information. Motivated by this, this article proposes a federated recommendation model by integrating both social relationships and KG. However, directly applying these two strategies to FedRS could lead to privacy leakage and increased communication overhead. To address this, this work encrypts user embeddings of clients using homomorphic encryption before uploading them to the server and adds pseudo-friends and pseudo-items when uploading user social and interaction lists to prevent privacy leakage. Then, the server distributes friend embeddings and pretrained entity and relationship embeddings to clients based on the social and interaction lists, reducing communication overhead. Finally, clients employ the graph attention network to integrate social relationships and KG to improve the accuracy of the recommendation. Experiments on two real datasets demonstrate that our federated recommendation model achieves improvements of 1.8%, 2.1% in area under curve and 3.3%, 1.9% in F1 score over the last federated recommendation methods.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6502-6516"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097451","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Transfer Learning for Cross-Domain Sentiment Classification via Domain Adaptation 基于领域自适应的跨领域情感分类迁移学习

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2025-11-21 DOI: 10.1109/TCSS.2025.3625989

Rahul Kumar Singh;Agampreet Saini;Erik Cambria;Abhijit Kumar

{"title":"Transfer Learning for Cross-Domain Sentiment Classification via Domain Adaptation","authors":"Rahul Kumar Singh;Agampreet Saini;Erik Cambria;Abhijit Kumar","doi":"10.1109/TCSS.2025.3625989","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3625989","url":null,"abstract":"The exponential proliferation of user-generated content across online platforms has driven a substantial surge in domain-specific textual data. Consequently, sentiment analysis has emerged to facilitate context-aware, real-time decision making in dynamic, data-intensive domains. Nevertheless, obtaining large-scale annotated data for all domains remains impractical, therefore a necessity for robust cross-domain sentiment classification techniques is still necessary. This work proposes WBiGe a hybrid domain adaptation architecture that combines static and contextual embeddings with a hierarchical BiGRU encoder to exploit semantic structure globally and sentiment cues locally. A lightweight convolutional architecture is used to extract phrase level patterns that are usually ignored by sequential encoders to improve short range sentiment expressions detection. To focus sentiment informative pivots and nonpivots during domain drilldown, a class aware domain alignment mechanism based on weighted maximum mean discrepancy is utilized. Furthermore, an iterative pseudolabeling strategy and dual-level attention modules are incorporated into the architecture to continuously improve target domain prediction to increase the interpretability of the model. The proposed WBiGe framework was extensively assessed using the Amazon benchmark dataset, in which it significantly outperforms the existing models across a wide range of cross-domain classification tasks with a record peak accuracy of 92.75%.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6609-6624"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097626","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A High-Efficiency Government Chatbot: Multisource-Aware Algorithm Oriented Conversation-Level Goal 一个高效的政府聊天机器人：面向会话级目标的多源感知算法

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2026-03-05 DOI: 10.1109/TCSS.2025.3633460

Zhixuan Lian;Fang Wang;Meiyin Huang

{"title":"A High-Efficiency Government Chatbot: Multisource-Aware Algorithm Oriented Conversation-Level Goal","authors":"Zhixuan Lian;Fang Wang;Meiyin Huang","doi":"10.1109/TCSS.2025.3633460","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3633460","url":null,"abstract":"Government chatbots (G-chatbots) are pivotal tools for improving public service delivery by enabling efficient and high-quality information exchange between governments and citizens. In practice, however, both efficiency and quality are undermined by persistent challenges in user understanding: chatbots often capture only short-term intentions while failing to infer long-term goals, leading to topic drift away from users’ core needs; moreover, conversations may extend into knowledge areas beyond system coverage, producing excessive dialogue turns that compromise efficiency and fragmented responses that erode quality, ultimately diminishing user patience. To address these issues, this study proposes a multisource-aware G-chatbot algorithm oriented toward conversation-level goals. The algorithm introduces two novel modules: 1) the multisource-aware (MSA) Adapter, which integrates user intentions, dialogue context, and historical knowledge to enhance semantic understanding across multiturn interactions; and 2) the virtual edge mechanism, which dynamically optimizes guidance paths by uncovering implicit relations in the knowledge graph, thereby reducing redundant exchanges and aligning responses with user objectives. Experimental results demonstrate consistent gains, including a 2% improvement in intent understanding, a 6% increase in answer quality, and a 1% reduction in dialogue turns, achieved with parameter-efficient fine-tuning (0.01%–13%). This research contributes to social intelligence and social systems modeling by offering a scalable solution for optimizing government–citizen interaction.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6682-6699"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097719","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Promoting Cooperation in Spatial Public Goods Games via Local Mean Field Proximal Policy Optimization With Neighbor-Dependent Cooperative Density Punishment 基于邻居依赖的合作密度惩罚的局部平均场近端政策优化促进空间公共物品博弈中的合作

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2026-08-12 DOI: 10.1109/TCSS.2026.3720953

Jinshuo Yang;Zhaoqilin Yang;Wenjie Zhou;Xin Wang;Youliang Tian

{"title":"Promoting Cooperation in Spatial Public Goods Games via Local Mean Field Proximal Policy Optimization With Neighbor-Dependent Cooperative Density Punishment","authors":"Jinshuo Yang;Zhaoqilin Yang;Wenjie Zhou;Xin Wang;Youliang Tian","doi":"10.1109/TCSS.2026.3720953","DOIUrl":"https://doi.org/10.1109/TCSS.2026.3720953","url":null,"abstract":"Spatial public goods games involve localized interactions and high-dimensional state spaces, making stable cooperation difficult. Traditional approaches often struggle to effectively capture neighborhood-level strategic interactions and dynamically align individual incentives with collective welfare. To resolve this issue, this article introduces a novel intelligent decision-making framework called local mean field proximal policy optimization with neighbor-dependent cooperative density punishment (LMFPPO-NCDP). The conventional mean field concept is reformulated as a localized statistical representation embedded into policy learning, enabling agents to adapt to neighborhood level cooperation dynamics. Additionally, a NCDP mechanism penalizes defectors according to local cooperation density without direct costs to cooperators. Experimental results demonstrate that LMFPPO-NCDP promotes rapid and stable cooperation under low enhancement factors. Compared with proximal policy optimization (PPO), Q-learning, and the Fermi update rule, the proposed framework achieves lower cooperation thresholds and more robust cooperative outcomes. Statistical analyses further confirm its robustness and effectiveness across different enhancement factors. The proposed framework provides a computational framework for analyzing adaptive cooperation in spatially structured multiagent systems.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"5491-5500"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097730","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Detection of Depression-Specific Circadian Misalignment in Young People via Whole-Day HRV and Actigraphy Patterns 通过全天HRV和活动图模式检测年轻人抑郁症特异性昼夜节律失调

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2025-12-31 DOI: 10.1109/TCSS.2025.3636524

Jilong Yan;Lu Jia;Miao Liu;Feng Qin;Guangyuan Liu;Wanhui Wen

{"title":"Detection of Depression-Specific Circadian Misalignment in Young People via Whole-Day HRV and Actigraphy Patterns","authors":"Jilong Yan;Lu Jia;Miao Liu;Feng Qin;Guangyuan Liu;Wanhui Wen","doi":"10.1109/TCSS.2025.3636524","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3636524","url":null,"abstract":"Circadian misalignment has been increasingly associated with depressive symptoms in young individuals, contributing to the persistence and recurrence of depression. Timely detection of such misalignment is therefore critical for early intervention. This study investigates circadian rhythm disturbances in physiological responses linked to depression and proposes a computational framework for symptom visualization and automated detection. A total of 47 features were extracted from electrocardiogram, photoplethysmography and tri-axial accelerometer signals collected over 1084 days from 148 participants. Daily recordings were divided into nonoverlapping time slots, resulting in a 47 × 153 feature matrix per individual per day. We introduced a novel index, the parasympathetic activation index (PAI), to quantify circadian variation in parasympathetic activity and to capture depression-related disruptions. Statistical analyses indicate that individuals with depressive symptoms exhibit significant circadian misalignment, characterized by elevated nighttime PAI and reduced daytime PAI. A deep neural network model trained on this feature representation achieved an F1-score of 85% under a leave-one-subject-out cross-test scheme. These results highlight the feasibility of leveraging wearable physiological signals and machine learning for real-world, user-independent depression monitoring and early risk identification in young populations.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6557-6570"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097769","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Textless, Tiny, and Private: Hashed n-Grams With LLM-Teacher Distillation and Local Differential Privacy for Mental-Health Screening on Social Media 无文本，微小，私密：哈希n-Grams与LLM-Teacher蒸馏和局部差异隐私在社交媒体上的心理健康筛查

IF 4.6 2区计算机科学

IEEE Transactions on Computational Social Systems Pub Date : 2026-10-01 Epub Date: 2026-03-25 DOI: 10.1109/TCSS.2026.3693692

Dandan Liu;Guangrui Fan;Qian Guo;Rui Zhang;Lihu Pan

{"title":"Textless, Tiny, and Private: Hashed n-Grams With LLM-Teacher Distillation and Local Differential Privacy for Mental-Health Screening on Social Media","authors":"Dandan Liu;Guangrui Fan;Qian Guo;Rui Zhang;Lihu Pan","doi":"10.1109/TCSS.2026.3693692","DOIUrl":"https://doi.org/10.1109/TCSS.2026.3693692","url":null,"abstract":"We present a textless, compute-light pipeline for user-level mental-health screening on social media that combines salted hashed character <inline-formula><tex-math>$n$</tex-math></inline-formula>-grams, local differential privacy (LDP), and large language model (LLM)-teacher concept distillation into a tiny, calibrated logistic classifier. The encoder stores no vocabulary or raw text; instead, all posts for a user are aggregated on device into a salted hashed presence vector, and a subsampled randomized-response mechanism is applied once per user to provide an explicit user-level LDP guarantee for this representation. To recover interpretability without online LLMs, we invoke frozen LLMs (one for Chinese and one for English) on a small, stratified subset to label clinically grounded concepts (e.g., hopelessness, self-blame, sleep disturbance, and help-seeking); compact students are then trained on hashed features to predict these concepts across the full corpus. On two large Weibo user-level corpora, our Hashed + differential privacy (DP ) + Concepts model achieves calibrated utility competitive with a dense-embedding baseline while retaining a CPU-only footprint. Relative to a hashed baseline without concepts, we observe consistent gains in macro-AUPRC and substantially lower calibration error after isotonic mapping. Privacy audits show that, at our default user-level privacy budget, LDP plus salt rotation reduces user-level membership-inference attack success, resists reconstruction compared to embedding inversion, and substantially lowers cross-deployment linkage area under the curve (AUC) toward chance in our empirical setting. We release code and audit scripts to facilitate responsible reuse.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 5","pages":"6758-6772"},"PeriodicalIF":4.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097863","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0