User Profiles Research Articles

Automated Extraction of Structured Data from the Social Network Instagram

The paper explores the extraction of structured information from the social network Instagram through a suitable application programming interface, namely the unofficial Instagram Private API. It focuses on creating a computer program that identifies which posts a user has tagged as "Likes" and then stores this information for profiling specific user profiles. The introduction of the paper highlights the general use of social media in modern society and the importance of personal data for these platforms. It specifies the aim of the study, which is to extract information from Instagram and then analyse it for user profiling. It then describes the evolution of the social network Instagram and key features such as different types of posts. This paper further focuses on the solution and implementation by using Python programming language to minimize the load on Instagram servers and reduce the risk of detection of automated processes. It describes the process of setting up new Instagram accounts, the obstacles in obtaining login credentials, and the need to simulate human behaviour to bypass the network's defence mechanisms. It then focuses on the actual retrieval of information such as the users followed, their posts and information about which posts the user has marked as favourites. It mentions that extracting data from closed profiles is difficult and elaborates on the technical challenges associated with this task. A significant part of this paper is a discussion of Instagram's defence mechanisms that respond to automated computer programs. It describes access denial, account blocking, and identity verification prompts such as CAPTCHA tests. Finally, the conclusion summarizes the results obtained, which indicate the acquisition of approximately 90,000 records for user profiling. It discusses the shortcomings of a fully automated solution due to Instagram's account creation conditions and defence mechanisms. It mentions the need for further research and highlights key gaps and challenges in this area. Overall, the study highlights the technical and security challenges in extracting information from Instagram and emphasises the need for further research and improvements in the technical procedures for extracting data from the platform.

Enhancing user engagement and satisfaction through personalized news recommendation systems

Personalized news recommendation systems have emerged as essential tools in addressing information overload by tailoring news content to individual user preferences. This paper provides a comprehensive overview of the advanced techniques employed in these systems, their impacts on user engagement, and the ethical considerations surrounding their development and implementation. We delve into the intricacies of data collection, processing, and user profiling, highlighting the methodologies and challenges inherent in each stage. Additionally, we explore advanced algorithmic foundations, including collaborative filtering, content-based filtering, and hybrid methods, elucidating their strengths and limitations. Furthermore, we examine the dynamics of user engagement within personalized news recommendation systems, analyzing key metrics and the role of user feedback in refining recommendation algorithms. Finally, we address privacy concerns, data sparsity issues, and biases, proposing solutions to mitigate ethical challenges and uphold user trust and fairness. This paper serves as a comprehensive guide for researchers, practitioners, and policymakers navigating the complexities of personalized news recommendation systems.

Brain-Computer Interface Integration With Extended Reality (XR): Future, Privacy And Security Outlook

The Brain-Computer Interface (BCI) is a rapidly evolving technology set to revolutionize our perception of the Internet of Things (IoT). BCI facilitates direct communication between the brain and external devices, enabling the control or interaction of devices without physical intervention. BCI technology is becoming more sophisticated, allowing third-party software embedded in emerging technologies such as Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR) to access sensors that read brain activity. These can be grouped under the umbrella term Extended Reality (XR). While BCI technology is disrupting the way data is collected, interpreted, and utilized within IoT networks, it is important to consider the potential privacy and security threats that it poses. Previous and not-so-recent cybersecurity research only scratched the surface in terms of security and privacy aspects of the then-emerging neural and brain-connecting technologies. However, recent advances in reconstructing language, music tracks, and imagery solely based on decoding neural signals pose a significant risk of mental privacy invasion and cybersecurity abuse. In this paper, we present an analysis of the potential threats posed by the integration of BCI with VR, AR, and MR. We analyze the involvement of major technological players in shaping BCI and XR advancements, examining the potential for these technologies to create detailed user profiles and reshape the monetization of user data in the ever-more-aggressive data-driven economy. We also outline a position view on the cybersecurity aspects that are not related to privacy and profiling per se, for example, cybersecurity attacks on the brain (e.g., ``brain rewriting'' attacks) facilitated by potentially vulnerable XR-BCI devices and software. The paper concludes by emphasizing the need for further research on the privacy and security implications of XR-BCI integration and inviting deeper exploration of the topic beyond theoretical papers and toward a more applied experimental setup.

Real-time V2G simulator for grid system frequency and voltage regulation with fault ride-through capabilities

Electric vehicles (EVs) have become one of the promising power grid ancillary services. They can help reduce the intermittent of renewable energy sources by supporting active power and frequency control services. Additionally, bi-directional chargers included into EV batteries can supply/consume reactive power for voltage regulation. To improve voltage quality, a unique control approach is devised to offer fault ride-through. In this study, real-time charging optimization schemes based on particle swarm optimization are developed to coordinate the EVs charging/discharging power with the grid power dispatches depending on the charging priority and preferential electricity prices. Five different car user profiles including different initial state of charge and plug in state have been depicted. Different optimum issues are carried out with comprehensive simulations employing hardware in the loop via real-time simulation for a modified IEEE 14-bus test system. Where different frequency and voltage controls, management strategies, variant fault scenarios, benefits, and challenges for power system regulation with EVs are illustrated. According to the simulation results, the proposed optimized charging strategies reduce the total expenses of EV parking station by 40 % compared to other cases, hence maximizing EV owner satisfaction. The net load curve is smoothed and reduced by 11 %, particularly during peak period.

Psychotherapeutische Sprechstunde am Arbeitsplatz: Zusammenhänge zwischen Unternehmensgröße und psychosomatischer Gesundheit

ZusammenfassungHintergrundDie Inanspruchnahme der psychotherapeutischen Sprechstunde am Arbeitsplatz (PT-A) wurde bislang in Großunternehmen (GU) untersucht. Diese unterscheiden sich strukturell von mittleren und Klein(st)unternehmen (KMU). Unterschiede der Nutzerprofile einer PT‑A hinsichtlich psychosomatischer Gesundheit, arbeitsbezogener Selbstwirksamkeit und Arbeitsfähigkeit sowie des psychosozialen Sicherheitsklimas (PSC) abhängig von der Unternehmensgröße wurden bisher kaum betrachtet.MethodenIn der Interventionsstudie „Frühe Intervention am Arbeitsplatz“ (friaa) wurden zwischen 09/2021 und 01/2023 an einer PT‑A interessierte Beschäftigte aus GU und KMU deutschlandweit befragt. Mittels t‑ und χ2-Tests wurden Unterschiede zwischen Beschäftigten in GU (n = 439) und KMU (n = 109) hinsichtlich F‑Diagnosen nach ICD-10 („International Statistical Classification of Diseases and Related Health Problems“; psychische und Verhaltensstörungen), Depressivität (PHQ-9), Ängstlichkeit (GAD-2), allgemeinen Funktionsniveaus (GAF), somatischer Symptombelastung (SSS-8), Gesundheit (VR-12), Arbeitsfähigkeit (WAI), Selbstwirksamkeit (SOSES) und psychosozialen Sicherheitsklimas (PSC-4) geprüft und mittels Korrelationsanalyse explorativ deren Zusammenhänge untersucht.ErgebnisseBeide Gruppen zeigten sich ähnlich stark beansprucht. In GU wurden aus Sicht der Beschäftigten psychosoziale Themen signifikant häufiger thematisiert als in KMU mit einer mittleren Effektgröße. Die Studie lieferte erste Hinweise, dass in GU positive Zusammenhänge des PSC‑4 mit SOSES und WAI sowie negative mit PHQ‑9 und SSS‑8 vorliegen.DiskussionDie vergleichbare psychische Beanspruchung der Beschäftigten in GU und KMU weist auf den Bedarf von verhältnis- und verhaltenspräventiven Maßnahmen unabhängig von der Unternehmensgröße hin. Vor allem in KMU sollte die Thematisierung psychosozialer Gesundheit einen größeren Stellenwert einnehmen.

Hierarchical representation learning for next basket recommendation

The task of next basket recommendation is pivotal for recommender systems. It involves predicting user actions, such as the next product purchase or movie selection, by exploring sequential purchase behavior and integrating users’ general preferences. These elements may converge and influence users’ subsequent choices. The challenge intensifies with the presence of varied user purchase sequences in the training set, as indiscriminate incorporation of these sequences can introduce superfluous noise. In response to these challenges, we propose an innovative approach: the Selective Hierarchical Representation Model (SHRM). This model effectively integrates transactional data and user profiles to discern both sequential purchase transactions and general user preferences. The SHRM’s adaptability, particularly in employing nonlinear aggregation operations on user representations, enables it to model complex interactions among various influencing factors. Notably, the SHRM employs a Recurrent Neural Network (RNN) to capture extended dependencies in recent purchasing activities. Moreover, it incorporates an innovative sequence similarity task, grounded in a k-plet sampling strategy. This strategy clusters similar sequences, significantly mitigating the learning process’s noise impact. Through empirical validation on three diverse real-world datasets, we demonstrate that our model consistently surpasses leading benchmarks across various evaluation metrics, establishing a new standard in next-basket recommendation.

Designing UI/UX for Mobile Learning on Programming Language Material Using the Design Thinking

The rapid advancement of technology in the Industry 4.0 revolution has driven organizations to adapt through digital transformation, including in the field of education, allowing for more efficient and effective learning processes in line with the needs of society5.0. This article discusses the UI/UX design of a mobile learning application for programming language material using the Design Thinking method, which includes five stages: Empathize, Define, Ideate, Prototype, and Testing, with a focus on user understanding and needs. The design process involves creating prototypes and testing to ensure a user-friendly design. The main features of the application include a splash screen, welcome page, login and sign-up pages, home page, material pages, notes, user profile, and an Executor feature for writing and running code directly. Testing results using the System Usability Scale (SUS) showed an average score of 65, falling into the "adequate" category, indicating that the design functions well. However, further research is needed to refine this application, highlighting the importance of a structured and user-focused approach in designing digital educational applications.

OurSCARA: Awareness-Based Recommendation Services for Sustainable Tourism

Sustainable tourism has emerged as a critical concern in contemporary society due to its potential to mitigate negative environmental and socio-cultural impacts associated with traditional tourism practices. In this context, recommendation systems (RS) are crucial in shaping travelers’ choices toward sustainable options. This research article proposes an innovative approach to RS tailored for sustainable tourism, termed Sustainability and Cultural Awareness-based Recommendation Algorithm (OurSCARA). OurSCARA integrates awareness of environmental and socio-cultural factors (sustainability attributes) into the recommendation process to facilitate informed decision-making by travelers. The system leverages data analytics techniques, including sentiment analysis, user profiling, and collaborative filtering (CF), to personalize recommendations based on users’ preferences, sustainability preferences, and contextual factors. Furthermore, OurSCARA incorporates real-time data sources such as weather conditions, local events, and community initiatives to enhance the relevance and timeliness of recommendations. A prototype implementation of OurSCARA is presented, along with a comprehensive evaluation framework to assess its effectiveness in promoting sustainable tourism behaviors. Through empirical evaluation using datasets collected from TripAdivsor, we demonstrate the potential of OurSCARA to influence traveler behavior towards more sustainable choices while enhancing their overall tourism experience. The findings underscore the significance of integrating sustainability considerations into RS and pave the way for future research and development in this emerging area at the intersection of computer science and sustainable tourism.

Optimizing community green roof spaces in high-density cities: a K-modes clustering algorithm based analysis of resident preferences and spatial configuration

ABSTRACT Green roofs are an important part of the ecological system in high-density cities and a crucial component of community public green spaces. Analyzing preferences of residents during spatial planning can help improve their satisfaction and efficiency in using community rooftop spaces. This study uses K-modes algorithm to perform cluster analysis on questionnaires from 699 residents, and summarizes user profiles with significant characteristics by combing the comprehensive ratings of residents regarding different roof space functions. The study shows that differences in preferences among people are not only related to demographic characteristics but also to their interests and environmental perceptions. For example, low-to-middle-income groups, being price-sensitive, tend to reduce energy consumption expenditure. The highly educated population, driven by social needs, shows a clear preference for activity spaces and sports facilities. The elderly population, emphasizing healthy eating and rural memories, gives higher ratings to agricultural spaces. Finally, this study explores the correlation between certain resident characteristics and their preferences for green roof space functions, and proposes a “1+X” spatial configuration strategy dominated by Landscape Leisure Space, with Ecological Low-carbon Space, Agricultural Production Space, and Activity Gathering Space as options, in order to optimize community green roof spaces guided by resident needs.

Unlocking Interactive Learning: Building a Robust Quiz Portal with MERN Stack Technology"

Online learning has become increasingly popular in the digital era, necessitating innovative platforms to effectively engage students. This research outlines the development and implementation of an interactive quiz portal utilizing the MERN (MongoDB, Express.js, React.js, and Node.js) stack to enhance learning outcomes. The portal features a user-friendly interface for easy navigation and access to a diverse range of quizzes on various subjects. To ensure scalability, performance, and real-time interactions, Node.js handles backend operations, React.js manages dynamic interfaces, Express.js is used for server-side development, and MongoDB stores data. The portal includes instant feedback mechanisms, gamification elements, and personalized user profiles to enhance user motivation and involvement. Furthermore, adaptive algorithms are employed to adjust quiz content based on user proficiency levels. This paper presents the design, development, and implementation of an online test platform using the MERN (MongoDB, Express, React, and Node.js) stack. The portal functions as an all-inclusive platform for administering tests, assessments, and learning evaluations in an online setting. The portal provides smooth integration of frontend and backend activities, guaranteeing strong performance and user interaction, by utilizing the adaptability and scalability of the MERN stack. The portal enables quick development and deployment cycles with the help of MongoDB's document-based data storage, Express.js's effective middleware and routing, React.js's dynamic UI components, and Node.js's event-driven architecture. To improve user experience and learning results, features including interactive feedback mechanisms, real-time performance monitoring, quiz design and administration, and user authentication are easily integrated.

Profiles of cannabis users and impact on cannabis cessation.

Although cannabis was legalized in Canada in 2018 and is one of the most used substances in Canada, few studies have examined how individuals with different patterns of cannabis use differ in their attempts to decrease or abstain from cannabis. The current study examined how groups of cannabis users, which were formed on the basis of demographic characteristics, substance use patterns, mental health symptoms, and self-reported quality of life differed on their experiences with cannabis cessation. A sample of 147 Canadian adult participants who had attempted to decrease or quit cannabis were recruited from the community (n = 84, 57.14%) and crowdsourcing (n = 63, 42.86%). Four profiles of cannabis users emerged using a Latent Profile Analysis: low-risk (n = 62, 42.18%), rapidly escalating high-risk (n = 40, 27.21%), long-term high severity (n = 35, 23.81%), and long-term lower severity (n = 10, 6.80%). Individuals in the rapidly escalating profile had attempted to decrease their cannabis use more times compared to other profiles. More participants in the long-term high severity group found their use stayed the same or got worse after their last cessation attempt, compared to the low-risk group where more individuals indicated their use stopped. The results of the current study indicate that cannabis users differ in their attempts at reducing or ceasing cannabis use and that they may benefit from different intensity of cannabis interventions.

Simulation of E-learning video recommendation based on virtual reality environment on English teaching platform

E-learning video is an effective teaching tool, and virtual reality technology can enhance the learning experience. However, the current English teaching platform lacks the personalized and real-time feedback video recommendation system. Therefore, this study proposes an E-learning video recommendation system based on virtual reality environment, which aims to provide a personalized, interactive and real-time feedback learning experience. In this paper, the prototype of E-learning video recommendation system based on virtual reality environment is designed, and the simulation experiment is carried out on the English teaching platform. In the environment of the E-learning platform, human–computer interaction and entertainment media can enhance students’ learning process and assist in teaching work. This article analyzes the simulation of E-learning video recommendation based on virtual reality environment on English teaching platforms. The E-learning video recommendation system can create user profiles on the website based on their basic information and behavior, fully tap into their preferences, and provide personalized learning services for them. The recommendation system can be divided into two parts: offline and online. Offline systems are mainly responsible for creating video related content, while online systems store data in databases, allowing for real-time investigation and retrieval of data; Online systems can respond to user requests in real-time, and combining online extraction with user behavior analysis can generate the final recommendation results. While constructing an artificial intelligence platform for English teaching, it is also necessary to propose application strategies for English teaching and achieve modernization of curriculum teaching.

E-commerce Platform in Digital Economy: Analysis and Optimization of Amazon's Footwear Store

With the continuous deepening of globalization, cross-border e-commerce has become an important part of global trade. Especially driven by digital payment technology, network technology, industrial upgrading and policy support, China's cross-border e-commerce industry has developed rapidly. This research aims to analyze the sales data of men's footwear on the Amazon platform, understand market trends, user behavior patterns, and product characteristics, and then provide data-supported marketing strategies and operational decisions for e-commerce platforms. We have adopted a variety of data analysis techniques, including data cleaning, ARIMA model prediction, price band analysis, repeat purchase rate analysis, user profiling (RFM model), sentiment analysis and cluster analysis. Through comprehensive analysis of multi-dimensional information such as store sales data, product attributes, market hot words, and user evaluations, market dynamics and user preferences are revealed. We have comprehensively applied a variety of data analysis models and techniques, such as using the ARIMA model for sales prediction, the LDA model for sentiment analysis, and the RFM model combined with cluster analysis to construct a refined user profile. Through in-depth analysis of user transaction behaviors and product attributes, accurate marketing strategies can be provided for e-commerce platforms, enhance user experience, increase user stickiness, and thus achieve business growth and improvement of profitability. At the same time, the research also points out the importance of brand building, supply chain management, risk assessment and other aspects, providing strategic suggestions for the long-term stable development of cross-border e-commerce.

Version control of speaker recognition systems

This paper discusses one of the most challenging practical engineering problems in speaker recognition systems — the version control of models and user profiles. A typical speaker recognition system consists of two stages: the enrollment stage, where a profile is generated from user-provided enrollment audio; and the runtime stage, where the voice identity of the runtime audio is compared against the stored profiles. As technology advances, the speaker recognition system needs to be updated for better performance. However, if the stored user profiles are not updated accordingly, version mismatch will result in meaningless recognition results. In this paper, we describe different version control strategies for speaker recognition systems that had been carefully studied at Google from years of engineering practice. These strategies are categorized into three groups according to how they are deployed in the production environment: device-side deployment, server-side deployment, and hybrid deployment. To compare different strategies with quantitative metrics under various network configurations, we present SpeakerVerSim, an easily-extensible Python-based simulation framework for different server-side deployment strategies of speaker recognition systems.

PWD’S Learning App

The goal of the proposed project is to create a mobile application specifically designed to meet the educational needs of people with impairments, with an emphasis on pupils who are blind or deaf. The program consists of two modules: a Basics of Coding module for deaf users and an English Learning module for blind users. The English module includes gamification features like levels and streak systems to increase student engagement, a voice-controlled interface to enhance interaction, and customized learning algorithms to modify the complexity of the lessons based on performance. In contrast, the Coding module makes use of a visual user interface that is accessible to students who are deaf. It does this by teaching coding principles through visually appealing material and by integrating game-like elements and individualized learning algorithms. Prioritizing inclusiveness and accessibility, both modules guarantee a flawless mobile user experience. To improve the learning process, additional features like progress monitoring and user profiles are included. Based on user input, the program will be updated and improved continuously to ensure that it continues to be relevant and successful in fostering diversity and skill development in education. Keywords— PWD, Gamified Education, Hearing Listening and Blind problem.

Cannabis use, mental health, and problematic Internet use in Quebec: A study protocol.

Problematic Internet use is characterized by excessive use of online platforms that can result in social isolation, family problems, psychological distress, and even suicide. Problematic Internet use has been associated with cannabis use disorder, however knowledge on the adult population remains limited. In Quebec, cannabis use has significatively increased since 2018, and it is associated with various risks in public safety, public health, and mental health. This study aims to identify factors associated with problematic Internet use among adult cannabis users and to better understand their experiences. This project is a mixed explanatory sequential study consisting of two phases. Phase 1 (n = 1500) will be a cross-sectional correlational study using probability sampling to examine variables that predispose individuals to problematic Internet use, characteristics associated with cannabis use, Internet use, and the mental health profile of adult cannabis users in Quebec. Descriptive analyses and regression models will be used to determine the relationship between cannabis use and Internet use. Phase 2 (n = 45) will be a descriptive qualitative study in the form of semi-structured interviews aimed at better understanding the experience and background of cannabis users with probable problematic Internet use. The results of this study will support the development of public policies and interventions for the targeted population, by formulating courses of action that contribute to the prevention and reduction of harms associated with cannabis use and problematic Internet use. Furthermore, an integrated knowledge mobilization plan will aid in the large-scale dissemination of information that can result useful to decision-makers, practitioners, members of the scientific community, and the general population regarding the use of cannabis and the Internet.

A Privacy-Preserving Friend Matching Scheme Based on Attribute Encryption in Mobile Social Networks

In mobile social networks, users can easily communicate with others through smart devices. Therefore, the protection of user privacy in social networks is becoming a significant subject. To solve this problem, this paper proposes a fine-grained data access control scheme that uses attributes to match friends. In our scheme, the friend-making parties generate friend preference and self-description lists, respectively, realizing attribute hiding by converting friendship preference into ciphertext access control policies and self-description into attribute keys. The social platform matches user profiles to quickly eliminate unmatched users and avoids invalid decryption. In order to reduce the computational burden and communication cost of mobile devices, we adopt an algorithm mechanism for outsourcing decryption. When the user meets the matching conditions, the algorithm outsources the bilinear pair operation with large computation to the friend server. After that, the user finally decrypts the ciphertext. Security analysis shows that our scheme is safe and effective. In addition, performance evaluation shows that the proposed scheme is efficient and practical.

Scalable Career & Educational Growth Transitions: Developing a Mentoring Network & Engaging with an Online Mentoring Platform.

The National Research Mentoring Network's (NRMN) goal is to increase diversity and provide mentoring, networking, and professional development resources to underrepresented populations in STEMM. The virtual, scalable MyNRMN platform includes 15,136 mentees, 7,998 mentors, 3,937 institutions, and 11,832 connections. Our goal was to understand mentee growth and "transitions" alongside engagement with mentoring and networking resources in MyNRMN. In this study, we define transition as the act of a person making a change to their MyNRMN profile that includes one of the following profile fields: institution, education level, degree, or career. We hypothesized that diverse networks and engagement in MyNRMN would be significantly associated with positive transitions. We assessed 7,209 user profiles that indicated an education/career change during the study period. Connections between racially homogenous users were associated with decreased positive transitions (b=-0.107, p<0.05). Connections with heterogeneous ethnicities were associated with increased positive transitions (b=0.105, p<0.05). Longevity within MyNRMN platform was associated with increased positive transitions (b=0.424, p<0.05).

Bookhive: Virtual Society for Readers

In an era marked by technological advancements, the research endeavors to revolutionize the traditional paradigm of library management by introducing a comprehensive virtual society for readers enriched with digital features and user-centric functionalities. This system comprises multiple interconnected modules designed to enhance the overall library experience for both administrators and patrons. The core Library Management Module streamlines book cataloging, user management, and inventory tracking, optimizing the operational efficiency of libraries. Concurrently, the Mobile App Module extends the library’s reach beyond physical boundaries, offering a user-friendly platform for on-the-go access to resources, virtual book clubs, and personalized recommendations. The Community Engagement Module fosters a vibrant literary community by facilitating discussions, author-reader interactions, and collaborative book recommendations. The Delivery and Rare Books Module ensures accessibility by enabling users to request book deliveries and notifying them of rare book availability. Furthermore, the User Profile and Customization Module empowers patrons to personalize their library experience, from user registration to preference settings. This research is not merely a technological overhaul but a gateway to a more dynamic and inclusive library ecosystem, embracing the digital age while preserving the essence of community, engagement, and the joy of reading. With data-driven insights, personalized features, and an emphasis on community building, our Library Management System represents a forward-looking approach to redefining the role of libraries in the modern world, ensuring they remain vibrant hubs of knowledge, connection, and literary exploration.

Graph Neural Networks for Recommendation Leveraging Multimodal Information

Recommender systems act as filtering algorithms to provide users with items that might meet their interests according to the expressed preferences and items' characteristics. As of today, the collaborative filtering paradigm, along with deep learning techniques to learn high-quality users' and items' representations, constitute the de facto standard for personalized recommendation, showing remarkable recommendation accuracy performance. Nevertheless, recommendation remains a highly-challenging task. Among the most debated open issues in the community, this thesis considers two algorithmic and conceptual ones, namely: (i) the inexplicable nature of users' preferences, especially when they come in the form of implicit feedback; (ii) the effective exploitation of the collaborative information in the designing and training of recommendation models. In domains such as fashion, food, and media content recommendation, the shallow item's profile can be enhanced through the multimodal characteristics describing items [Malitesta et al., 2023]. Driven by these assumptions, in the first part of this thesis, we apply multimodal deep learning strategies for multimedia recommendation; the scope is to study and design recommendation algorithms based upon the principles of multimodality to possibly match each item's characteristic to the implicit preference expressed by the user [Deldjoo et al., 2022], thus addressing the (i) issue. Recent collaborative filtering approaches profile users and items through embedding vectors in the latent space. However, such models disregard structural properties naturally encoded into the user-item interaction data. Indeed, recommendation datasets are easily describable under the topology of a bipartite and undirected graph, with users and items being the graph nodes connected at multiple distance hops. In this respect, the application of graph neural networks , recent deep learning techniques specifically tailored to learn from non-euclidean data, can provide a refined representation of users and items to mine near- and long-distance relationships in the user-item graphs [Anelli et al., 2023b]. Indeed, this is one possible solution to exploit the collaborative information, which is effectively propagated within the user-item graph, addressing the (ii) issue. Conclusively, this thesis aims to match the two families of recommendation strategies by leveraging graph neural networks and multimodal information data [Anelli et al., 2022]. In doing so, other numerous micro-aspects within the two macro-areas (introduced above) are examined. Indeed, the thesis is a systematic compendium of careful analyses regarding, among others, reproducibility, novel evaluation dimensions [Anelli et al., 2023a], and tasks/scenarios complementary to recommendation. Awarded by : Politecnico di Bari, Bari, Italy on 30 January 2024. Supervised by : Tommaso Di Noia. Available at : https://hdl.handle.net/11589/264941.