Abstract In the digital landscape, social media has emerged as a prevalent channel for global communication, connecting like-minded individuals worldwide. However, while facilitating information exchange, it is also susceptible to the dissemination of false information, posing a constant challenge to the reliability of online content. To address this issue, this paper introduces a novel methodology called TM-FID (Topic-oriented Multimodal False Information Detection), which combines false information detection and neural topic modeling within a semi-supervised multimodal approach. By jointly leveraging textual and visual information contained in online news, our approach provides insights into how false information influences specific discussion topics, thus enabling a comprehensive and fine-grained understanding of its spread and impact on social media conversation. Experimental evaluation carried out on a set of multimodal gossip-related news demonstrates the quality of the identified topics, assessed through a novel centroid-based metric, as well as the efficacy of the cross-attention mechanism used within TM-FID to accurately identify false information in multimodal news. Overall, the proposed methodology can enable effective strategies to counter the spread of false information, thereby fostering trust and confidence in the information shared on social media platforms.

In the era of online learning, automated dynamic suggestions for fruitful collaborations among learners have become crucial. We propose an algorithm that takes into account ethical considerations and suggests interactions among students for collaborative involvement in group activities in an inherently fair manner: all students participate in fairly weighted activities and have a balanced number of weighted interactions. Students' preferences are taken into consideration and can be expressed either directly via their statements or indirectly via their conversations in the context of learning gamification. A synthetic preference graph that simulates students' views and thoughts is constructed. Moreover, we proceed to build a model for personalized interactions that prevents user isolation. The quality of the evolution of interactions is measured based on the diversity of interactions, the interaction graph edit similarity that affects storage cost, granularity of graph construction and fairness criteria. We use deep machine learning for sentiment detection in dialogues in order to conduct sentiment analysis of real data in the context of an online game chat among students. We categorize text as positive, negative or neutral in terms of emotion and produce student profiles. We produce embeddings with an accuracy of 92%. This analysis is extended to all texts in the discussion, resulting in a preference profile for each learner, from which ultimately a quantitative estimate of potential interaction derives for use in the interaction graph. Our approach can be utilised in the teaching practice as an efficient novel framework for online learning activities and gamification, and serves as an automated dynamic method for fair collaboration choices in teaching practice.

A significantly large proportion of HTTP traffic results from bandwidth-intensive multimedia content circulating through online social networks (OSNs). With multimedia content providers, such as YouTube, often relying on content delivery network (CDN) infrastructures, the pursuit lies in exploiting the user activity extracted from OSNs to improve the content prefetching mechanism. Aiming to reduce bandwidth usage, we incorporated a dynamic mechanism to CDNsim, a stand-alone CDN traffic simulator. This mechanism is based on a dynamic policy that takes patterns of information transmission over OSNs into account. Herein, we demonstrate that the performance of CDNs can be improved, and the cost of copying to surrogate servers is taken into consideration.