ABSTRACT This research explores the social response to disclosures and conversations about mental health on social media, which is a pioneering and innovative approach. Unlike previous studies, which focused predominantly on psychopathological aspects, this study explores how communities react to conversations about mental health on Instagram, one of the favourite social media platforms among young people, breaking new ground not only in the Spanish context, but also on a global scale, filling a gap in international research. The study created a novel corpus by collecting and labelling comments on Instagram posts related to celebrity mental health disclosures, categorising them by polarity (positive, negative, neutral) and stigma. Additionally, the research implements machine learning algorithms to detect stigma and polarity in mental health disclosures on Instagram. While traditional techniques like Support Vector Machine (SVM) and RF (Random Forest) displayed decent performance with lower computational loads, advanced deep learning and BERT (Bidirectional Encoder Representation from Transformers) algorithms achieved outstanding results. In fact, BERT models achieve around 96% accuracy in polarity and stigma detection, while deep learning models achieve 80% for polarity and 87% for stigma, very high accuracy metrics. This research contributes significantly to understanding the impact of mental health discussions on social media, offering insights that can reduce stigma and raise awareness. Artificial intelligence can be used for more responsible use of social media and effective management of mental health problems in digital environments.

The research addresses challenges related to mental health issues in social media by integrating natural language processing. First, the study extends a previous corpus labelled with emotions and polarity by including new Instagram and TikToK posts related to celebrity and influencer disclosures about mental health. This corpus is the first Spanish corpus designed to analyse the impact of social responses to mental health narratives on two of the most widely used social networks. Secondly, the research integrates BERT (Bidirectional Encoder Representations) classification models to improve emotion and polarity detection. One of the modelled algorithms, MenTaiBERT, leveraging a specialised classification layer demonstrates superiority over the other BERT algorithms, achieving 99% accuracy in emotion detection and 98% accuracy in polarity. Indeed, MenTaiBERT significantly outperforms the accuracy of the other algorithms by up to 13 percentage points. Third, A user-friendly graphical tool has been designed, based on the previous corpus and classification models, to help practitioners identify emotional patterns in social media posts related to mental health. In summary, analysing through innovative artificial intelligence strategies the emotional impact of celebrity posts on social networks is crucial, especially among young people, as these platforms significantly influence their self-esteem, perception of reality and emotional well-being.

This paper introduces a novel dynamic multipath routing, modulation level, spatial and spectrum assignment algorithm for space division multiplexing (SDM) enabled elastic optical networks (EON) with the aim of minimizing the blocking probability and the energy consumed by bandwidth variable transponders (BVTs). The adopted multipath routing strategy allows the splitting of the demand into several sublightpaths using different fiber cores but ensuring that all of them utilize the same set of fibers in order to avoid differential delay. The method also imposes continuity constraints in both spectrum and core location in order to use cost-effective SDM Reconfigurable Optical Add and Drop Multiplexers (ROADMs) without lane change support. The complete usage of multi-core fibers (MCFs) in this kind of networks is restricted due to inter-core crosstalk (XT), which can reduce the quality of received signals. Therefore, the method besides using the most effective modulation format, also ensures that the XT of the lightpaths (or sublightpaths) does not exceed the threshold for each modulation format. A simulation study comparing our method with another similar proposal from the literature is presented for different types of topologies in terms of link distances. Simulation results demonstrate that the proposed multipath routing algorithm in networks including links close to or beyond 1000 kms significantly boost the dynamic performance in terms of blocking probability, energy consumption, and latency.

This research explores for the first time the application of machine learning to detect emotional responses in video game streaming channels, specifically on Twitch, the most widely used platform for broadcasting content. Analyzing sentiment in gaming contexts is difficult due to the brevity of messages, the lack of context, and the use of informal language, which is exacerbated in the gaming environment by slang, abbreviations, memes, and jargon. First, a novel Spanish corpus was created from chat messages on Spanish video game Twitch channels, manually labeled for polarity and emotions. It is noteworthy as the first Spanish corpus for analyzing social responses on Twitch. Secondly, machine learning algorithms were used to classify polarity and emotions offering promising evaluations. The methodology followed in this work consists of three main steps: 1) Extracting Twitch chat messages from Spanish streamers’ channels related to gaming events and gameplays; 2) Processing and selecting the messages to form the corpus and manually annotating polarity and emotions; and 3) Applying machine learning models to detect polarity and emotions in the created corpus. The results have shown that a Bidirectional Encoder Representation from Transformers (BERT) based model excels with 78% accuracy in polarity detection, while deep learning and Random Forest models reach around 70%. For emotion detection, the BERT model performs best with 68%, followed by deep learning with 55%. It is worth noting that emotion detection is more challenging due to the subjective interpretation of emotions in the complex communicative context of video gaming on platforms such as Twitch. The use of supervised learning techniques, together with the rigorous corpus labeling process and the subsequent corpus pre-processing methodology, has helped to mitigate these challenges, and the algorithms have performed well. The main limitations of the research involve category and video game representation balance. Finally, it is important to stress that the integration of machine learning in video games and on Twitch is innovative, by allowing the identification of viewers’ emotions on streamers’ channels. This innovation could bring benefits such as a better understanding of audience sentiment, improving content and audience retention, providing personalized recommendations and detecting toxic behavior in chats.

In recent years, the number of embedded computing devices connected to the Internet has exponentially increased. At the same time, new applications are becoming more complex and computationally demanding, which can be a problem for devices, especially when they are battery powered. In this context, the concepts of computation offloading and edge computing, which allow applications to be fully or partially offloaded and executed on servers close to the devices in the network, have arisen and received increasing attention. Then, the design of algorithms to make the decision of which applications or tasks should be offloaded, and where to execute them, is crucial. One of the options that has been gaining momentum lately is the use of Reinforcement Learning (RL) and, in particular, Deep Reinforcement Learning (DRL), which enables learning optimal or near-optimal offloading policies adapted to each particular scenario. Although the use of RL techniques to solve the computation offloading problem in edge systems has been covered by some surveys, it has been done in a limited way. For example, some surveys have analysed the use of RL to solve various networking problems, with computation offloading being one of them, but not the primary focus. Other surveys, on the other hand, have reviewed techniques to solve the computation offloading problem, being RL just one of the approaches considered. To the best of our knowledge, this is the first survey that specifically focuses on the use of RL and DRL techniques for computation offloading in edge computing system. We present a comprehensive and detailed survey, where we analyse and classify the research papers in terms of use cases, network and edge computing architectures, objectives, RL algorithms, decision-making approaches, and time-varying characteristics considered in the analysed scenarios. In particular, we include a series of tables to help researchers identify relevant papers based on specific features, and analyse which scenarios and techniques are most frequently considered in the literature. Finally, this survey identifies a number of research challenges, future directions and areas for further study.

Mobile technologies have undergone a great leap forward in a few years, and while 5G networks are already being deployed, there are not yet many proven applications that can fully utilize the advantages of this new technology. Connected and autonomous vehicles are a specific and demanding case, particularly in terms of delay and bandwidth requirements, which can leverage not only 5G but also edge computing technologies. Therefore, the development of testbeds to demonstrate future applications is crucial to enable the full deployment of 5G and edge computing possibilities. In this paper, we present a flexible and modular testbed, targeted towards the evaluation of Cooperative, Connected, and Automated Mobility (CCAM) applications, and we demonstrate a use case (using a 4G system) where an autonomous vehicle offloads processing tasks to an edge server which analyzes images, makes routing decisions, and sends guidance commands back to the vehicle, thus proving the possibilities of edge computing and wireless technologies.

A novel quality of transmission (QoT) estimator based on support vector machines (SVM) is proposed for classifying optical connections (lightpaths) into high or low quality categories in impairment-aware wavelength-routed optical networks (WRONs). The performance of the SVM-based estimator is evaluated in a long haul communications network and compared to previous semi-analytical and cognitive proposals. Results show that the SVM approach significantly reduces the necessary computing time to estimate the QoT of a given lightpath, critical aspect of design in these networks, and even slightly improves accuracy.

The promises of SDN and NFV technologies to boost innovation and to reduce the time-to-market of new services is changing the way in which residential networks will be deployed, managed and maintained in the near future. New user-centric management models for residential networks combining SDN-based residential gateways and cloud technologies have already been proposed, providing flexibility and ease of deployment. Extending the scope of SDN technologies to optical access networks and bringing cloud technologies to the edge of the network enable the creation of advanced residential networks in which complex service function chains can be established to provide traffic differentiation. In this context, this paper defines a novel network management model based on a user-centric approach that allows residential users to define and control access network resources and the dynamic provision of traffic differentiation to fulfill QoS requirements.

An emulation study has been carried to demonstrate the benefit of a proactive restoration technique in cognitive heterogeneous optical networks. Results show the advantages of that method in terms of recovery percentage and disruption time.

Communication networks are facing continuous variations of traffic patterns as well as occasional failures of network equipment. Wavelength-routed optical networks (WRONs) offer the possibility of dynamically adapting to traffic conditions by means of reconfiguring the logical topology, that is, the set of lightpaths embedded in it. However, reconfiguration has a cost, the number of packets lost during the reconfiguration process. Hence, it is necessary to use efficient reconfiguration policies and algorithms to design the logical topologies. In this paper, a new algorithm is proposed to design logical topologies that jointly minimizes the number of lightpaths changed (reconfiguration cost) and the network congestion (reconfiguration reward). This method is based on the combination of genetic algorithms with Pareto optimality techniques. Thus, the algorithm provides a set of optimal (or near-optimal) solutions in terms of both parameters, the Pareto optimal set. Moreover, a novel policy to minimize the packet loss ratio considering all the solutions provided by the algorithm is also proposed. A simulation study is presented to show how the combination of the new algorithm and policy can reduce in more than one order of magnitude the packet loss ratio in stationary state and respond to abrupt changes in less time when compared with previous work on logical topology reconfiguration.