arXiv:2510.22904v1 Announce Type: cross 
Abstract: Social media has reshaped political discourse, offering politicians a platform for direct engagement while reinforcing polarization and ideological divides. This study introduces a novel topic evolution framework that integrates BERTopic-based topic modeling with Moral Foundations Theory (MFT) to analyze the longevity and moral dimensions of political topics in Twitter activity during the 117th U.S. Congress. We propose a methodology for tracking dynamic topic shifts over time and measuring their association with moral values and quantifying topic persistence. Our findings reveal that while overarching themes remain stable, granular topics tend to dissolve rapidly, limiting their long-term influence. Moreover, moral foundations play a critical role in topic longevity, with Care and Loyalty dominating durable topics, while partisan differences manifest in distinct moral framing strategies. This work contributes to the field of social network analysis and computational political discourse by offering a scalable, interpretable approach to understanding moral-driven topic evolution on social media.

تستكشف دراسة حديثة كيف غيرت وسائل التواصل الاجتماعي، وخاصة تويتر، الخطاب السياسي خلال الكونغرس الأمريكي السابع عشر. من خلال دمج نمذجة الموضوعات المستندة إلى BERTopic مع نظرية الأسس الأخلاقية، طور الباحثون إطارًا لتحليل تطور الموضوعات السياسية وأبعادها الأخلاقية. لا تسلط هذه المقاربة الضوء فقط على ديناميكيات المشاركة السياسية، بل تبرز أيضًا الاستقطاب المتزايد في المجتمع، مما يجعلها مساهمة مهمة في فهم التواصل السياسي الحديث.

Un estudio reciente explora cómo las redes sociales, especialmente Twitter, han transformado el discurso político durante el 117º Congreso de EE. UU. Al integrar la modelización de temas basada en BERTopic con la Teoría de Fundamentos Morales, los investigadores han desarrollado un marco para analizar la evolución y las implicaciones morales de los temas políticos. Este enfoque no solo ilumina las dinámicas del compromiso político, sino que también destaca la creciente polarización en la sociedad, lo que lo convierte en una contribución significativa para entender la comunicación política moderna.

Une étude récente examine comment les réseaux sociaux, en particulier Twitter, ont transformé le discours politique pendant la 117e Congrès des États-Unis. En intégrant la modélisation de sujets basée sur BERTopic avec la théorie des fondations morales, les chercheurs ont développé un cadre pour analyser l'évolution et les implications morales des sujets politiques. Cette approche éclaire non seulement les dynamiques de l'engagement politique, mais souligne également la polarisation croissante dans la société, ce qui en fait une contribution significative à la compréhension de la communication politique moderne.

A recent study explores how social media, particularly Twitter, has transformed political discourse during the 117th U.S. Congress. By integrating BERTopic-based topic modeling with Moral Foundations Theory, researchers have developed a framework to analyze the evolution and moral implications of political topics. This approach not only sheds light on the dynamics of political engagement but also highlights the growing polarization in society, making it a significant contribution to understanding modern political communication.

Modeling Political Discourse with Sentence-BERT and BERTopic

arXiv:2504.02800v3 Announce Type: replace 
Abstract: Mental disorders represent a critical global health challenge, and social media is increasingly viewed as a vital resource for real-time digital phenotyping and intervention. Large Language Models (LLMs) offer stronger semantic understanding and reasoning than traditional deep learning, but their use in high-stakes clinical settings is limited by hallucinations and the lack of persistent memory. However, existing literature has not sufficiently investigated how advanced enhancement techniques, specifically Retrieval-Augmented Generation (RAG) and Agentic systems, can address these reliability and reasoning limitations. Here, we systematically survey the evolving landscape of LLM-based methods for social media mental disorder analysis, spanning standard pretrained language models, RAG to mitigate hallucinations and contextual gaps, and agentic systems for autonomous reasoning and multi-step intervention. We organize existing work by technical paradigm and clinical target, extending beyond common internalizing disorders to include psychotic disorders and externalizing behaviors. Additionally, the paper comprehensively evaluates the performance of LLMs, including the impact of RAG, across various tasks. This work establishes a unified benchmark for the field, paving the way for the development of trustworthy, autonomous AI systems that can deliver precise and explainable mental health support.

أظهرت دراسة واستكشافات حديثة إمكانية استخدام نماذج اللغة الكبيرة (LLMs) في الكشف عن الاضطرابات النفسية من خلال وسائل التواصل الاجتماعي، مما يبرز أهمية تقنيات متقدمة مثل التوليد المعزز بالاسترجاع (RAG) والأنظمة الوكيلة لتحسين الموثوقية والتفكير في البيئات السريرية. تهدف هذه الأساليب إلى معالجة التحديات التي تطرحها الهلوسات وقيود الذاكرة في نماذج اللغة الكبيرة.

Una reciente encuesta y experimentos han destacado el potencial de los Modelos de Lenguaje Grande (LLMs) en la detección de trastornos mentales a través de las redes sociales, enfatizando la importancia de técnicas avanzadas como la Generación Aumentada por Recuperación (RAG) y sistemas agentes para mejorar la fiabilidad y el razonamiento en entornos clínicos. Estos métodos buscan abordar los desafíos planteados por las alucinaciones y las limitaciones de memoria en los LLMs.

Une récente enquête et des expériences ont mis en lumière le potentiel des grands modèles de langage (LLMs) dans la détection des troubles mentaux via les réseaux sociaux, soulignant l'importance de techniques avancées telles que la génération augmentée par récupération (RAG) et les systèmes agentiques pour améliorer la fiabilité et le raisonnement dans les contextes cliniques. Ces méthodes visent à relever les défis posés par les hallucinations et les limitations de mémoire des LLMs.

A recent survey and experiments have highlighted the potential of Large Language Models (LLMs) in detecting mental disorders through social media, emphasizing the importance of advanced techniques such as Retrieval-Augmented Generation (RAG) and Agentic systems to enhance reliability and reasoning in clinical settings. These methods aim to address the challenges posed by hallucinations and memory limitations in LLMs.

Survey and Experiments on Mental Disorder Detection via Social Media: From Large Language Models and RAG to Agents

arXiv:2512.08071v1 Announce Type: new 
Abstract: Crisis classification in social media aims to extract actionable disaster-related information from multimodal posts, which is a crucial task for enhancing situational awareness and facilitating timely emergency responses. However, the wide variation in crisis types makes achieving generalizable performance across unseen disasters a persistent challenge. Existing approaches primarily leverage deep learning to fuse textual and visual cues for crisis classification, achieving numerically plausible results under in-domain settings. However, they exhibit poor generalization across unseen crisis types because they 1. do not disentangle spurious and causal features, resulting in performance degradation under domain shift, and 2. fail to align heterogeneous modality representations within a shared space, which hinders the direct adaptation of established single-modality domain generalization (DG) techniques to the multimodal setting. To address these issues, we introduce a causality-guided multimodal domain generalization (MMDG) framework that combines adversarial disentanglement with unified representation learning for crisis classification. The adversarial objective encourages the model to disentangle and focus on domain-invariant causal features, leading to more generalizable classifications grounded in stable causal mechanisms. The unified representation aligns features from different modalities within a shared latent space, enabling single-modality DG strategies to be seamlessly extended to multimodal learning. Experiments on the different datasets demonstrate that our approach achieves the best performance in unseen disaster scenarios.

تقدم دراسة جديدة إطار عمل CAMO، الذي يستخدم التعميم العدائي متعدد النماذج الموجه بالسببية لتحسين تصنيف الأزمات من منشورات وسائل التواصل الاجتماعي. يهدف هذا النهج إلى تحسين استخراج المعلومات القابلة للتنفيذ المتعلقة بالكوارث، مع معالجة التحديات المتعلقة بالتعميم عبر أنواع الأزمات المختلفة.

Un nuevo estudio presenta el marco CAMO, que utiliza la generalización de dominio multimodal adversarial guiada por la causalidad para mejorar la clasificación de crisis a partir de publicaciones en redes sociales. Este enfoque busca mejorar la extracción de información útil relacionada con desastres, abordando los desafíos de la generalización a través de diversos tipos de crisis.

Une nouvelle étude présente le cadre CAMO, qui utilise la généralisation de domaine multimodale adversariale guidée par la causalité pour améliorer la classification des crises à partir des publications sur les réseaux sociaux. Cette approche vise à améliorer l'extraction d'informations exploitables liées aux catastrophes, en s'attaquant aux défis de la généralisation à travers divers types de crises.

A new study introduces the CAMO framework, which utilizes causality-guided adversarial multimodal domain generalization to enhance crisis classification from social media posts. This approach aims to improve the extraction of actionable disaster-related information, addressing the challenges of generalizing across diverse crisis types.

CAMO: Causality-Guided Adversarial Multimodal Domain Generalization for Crisis Classification

arXiv:2512.06293v1 Announce Type: new 
Abstract: Urban transit agencies increasingly turn to social media to monitor emerging service risks such as crowding, delays, and safety incidents, yet the signals of concern are sparse, short, and easily drowned by routine chatter. We address this challenge by jointly modeling linguistic interactions and user influence. First, we construct an influence-weighted keyword co-occurrence graph from cleaned posts so that socially impactful posts contributes proportionally to the underlying evidence. The core of our framework is a Poisson Deconvolution Factorization (PDF) that decomposes this graph into a low-rank topical structure and topic-localized residual interactions, producing an interpretable topic--keyword basis together with topic importance scores. A decorrelation regularizer \emph{promotes} distinct topics, and a lightweight optimization procedure ensures stable convergence under nonnegativity and normalization constraints. Finally, the number of topics is selected through a coherence-driven sweep that evaluates the quality and distinctness of the learned topics. On large-scale social streams, the proposed model achieves state-of-the-art topic coherence and strong diversity compared with leading baselines. The code and dataset are publicly available at https://github.com/pangjunbiao/Topic-Modeling_ITS.git

تستخدم وكالات النقل الحضري بشكل متزايد وسائل التواصل الاجتماعي لتحديد المخاطر الناشئة في الخدمة مثل الازدحام والتأخيرات والحوادث الأمنية. تم تطوير إطار عمل جديد يستخدم رسم بياني لتكرار الكلمات الرئيسية مع وزن التأثير وتحليل تفكيك بواسون لفحص المنشورات على وسائل التواصل الاجتماعي، مما يسمح باستخراج مواضيع قابلة للتفسير ودرجات أهميتها.

Las agencias de transporte urbano están utilizando cada vez más las redes sociales para identificar riesgos de servicio emergentes como la congestión, los retrasos y los incidentes de seguridad. Se ha desarrollado un nuevo marco que emplea un gráfico de co-ocurrencia de palabras clave ponderado por influencia y una Factorización de Deconvolución de Poisson para analizar publicaciones en redes sociales, lo que permite extraer temas interpretables y sus puntajes de importancia.

Les agences de transport urbain utilisent de plus en plus les réseaux sociaux pour identifier les risques de service émergents tels que la surpopulation, les retards et les incidents de sécurité. Un nouveau cadre a été développé qui utilise un graphique de co-occurrence de mots-clés pondéré par l'influence et une factorisation de déconvulsion de Poisson pour analyser les publications sur les réseaux sociaux, permettant d'extraire des sujets interprétables et leurs scores d'importance.

Urban transit agencies are increasingly utilizing social media to identify emerging service risks such as crowding, delays, and safety incidents. A new framework has been developed that employs an influence-weighted keyword co-occurrence graph and Poisson Deconvolution Factorization to analyze social media posts, allowing for the extraction of interpretable topics and their importance scores.

Modeling Political Discourse with Sentence-BERT and BERTopic

Was this article worth reading? Share it

Tweeteasy

MyFramework

Tribela