MTAT magistritööd – Master's theses

Selle kollektsiooni püsiv URIhttps://hdl.handle.net/10062/30974

Sirvi

Viimati lisatud

Nüüd näidatakse 1 - 20 1145
  • Kirje
    Secure Data Sharing in the Internet of Vehicles Using Blockchain-based Federated Learning
    (Tartu Ülikool, 2025) Luzan, Mykyta; Iqbal, Mubashar, juhendaja; Matulevičius, Raimundas, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    The Internet of Vehicles enables connected vehicles to share data and collaboratively learn to enhance road safety and traffic efficiency. Federated learning has emerged as a promising approach for enabling privacy-preserving collaborative learning among vehicles, allowing them to jointly train machine learning models without sharing raw sensitive data. However, the centralized architecture commonly used in federated learning introduces significant security vulnerabilities that can compromise system integrity and reliability. While extensive research exists on federated learning security in general, there is insufficient analysis of how these security challenges manifest in specific application contexts, particularly in dynamic environments like IoV. Here we show that integrating Hyperledger Fabric’s permissioned blockchain with zero-knowledge proofs creates a comprehensive security framework that effectively protects federated learning systems against both model tampering, aggregation protocol violation, and unauthorized access while maintaining privacy. Our systematic analysis and implementation reveals that blockchain technology can address core vulnerabilities in centralized federated learning architectures while preserving their privacy benefits, demonstrating advantages over previous approaches that relied solely on cryptographic protocols or trusted third parties. By validating our framework through a concrete IoV data sharing implementation, we establish a practical foundation for securing federated learning in distributed environments. The implications of this research extend beyond vehicular networks to any domain requiring secure collaborative learning among distributed participants. As autonomous systems become increasingly interconnected, this work demonstrates how combining blockchain with federated learning can enable trustworthy data sharing while preserving both privacy and security.
  • Kirje
    The Future of Digital Payments: Trends and Disruptions
    (Tartu Ülikool, 2025) Khasmammadov, Murad; Milani, Fredrik Payman, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    The evolution of digital payments have transformed the global financial ecosystem by providing faster, more secure, and more accessible ways to transact. However, despite the widespread adoption of digital payment systems, gaps remain in understanding the drivers of their evolution, the challenges that hinder their potential, and the implications of emerging trends such as blockchain, AI, and IoT. The rapid pace of technological advancement and regulatory changes underscore the need for a comprehensive overview of digital payment systems. This thesis addresses these gaps by exploring the evolution of digital payment systems, their lifecycle, and the various tools that drive this transformation. Using a systematic literature review, it addresses research questions related to the digital payment types, key drivers, benefits, challenges, and future trajectories of digital payments. The study highlights the role of technological advancements such as blockchain, AI, and IoT and the impact of regulatory frameworks in shaping payment innovations. Emerging trends, including cryptocurrencies, central bank digital currencies (CBDCs), and mobile payments, underscore the shift toward decentralization and financial inclusion. Despite significant progress, barriers such as interoperability, cybersecurity, and regulatory complexity persist. This thesis offers an understanding of the current landscape and future directions of digital payments, providing valuable insights for stakeholders, including consumers, businesses, financial institutions, and policymakers.
  • Kirje
    Real-time Pose Estimation of a Surgical Tool using Optical Coherence Tomography
    (Tartu Ülikool, 2025) Zaliznyi, Anton; Fishman, Dmytro, juhendaja; Kahrs, Lueder, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Minimally invasive and robotic-assisted surgeries have transformed medicine by reducing patient trauma, infection risk, and recovery times. Within these procedures, precise instrument tracking is critical, especially when navigating intricate anatomical structures and performing delicate interventions such as neurosurgery and microsurgery. Optical coherence tomography has emerged as a promising imaging modality that can provide high-resolution, real-time, and volumetric field-of-view for surgical sites. Existing methods that leverage optical coherence tomography for instrument pose tracking primarily focus on rigid instruments or rely on artificial markers; however, these approaches may fall short in practical scenarios involving occlusions and the dynamic nature of multi-jointed surgical tools. This thesis addressed this challenge by developing a markerless, high-speed, and accurate pose estimation method for an 8-degree-of-freedom microsurgical tool using optical coherence tomography. The proposed method achieves an average position error of 0.26 millimeters, an orientation error of 2.3 degrees, and joint angle errors of 1.9 and 1.9 degrees for θ1 and θ2, respectively, while operating with an inference speed of 20 milliseconds per volume. By eliminating the need for markers and being robust to occlusions, our method improves the reliability and feasibility of optical coherence tomography-based microsurgical instrument tracking in complex, dynamic, and realistic surgical environments. Future work should focus on testing this approach with more annotated real-world data and validating its effectiveness through in-vivo applications, thereby enhancing its reliability and practical impact.
  • Kirje
    Efficient Two-Party ML-DSA Protocol in Active Security Model
    (Tartu Ülikool, 2025) Kravtšenko, Semjon; Laud, Peeter, juhendaja; Krips, Toomas, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    ML-DSA is a NIST standard that defines a signature scheme: a set of algorithms for creating and verifying digital signatures. Digital signatures can be used, for example, to authenticate to websites online and to sign documents. ML-DSA signatures, unlike signatures that follow so-called classical formats, are quantum-resistant: it is believed that forging ML-DSA signatures is inviable even with a cryptographically relevant quantum computer (that is not yet known to exist). The security of a signing scheme relies on the secrecy of the used private key material. One way to increase the security of a signing scheme is to distribute the secret material across multiple devices, such that a sufficient number of them need to cooperate to create a signature. One scheme, that distributes the key across two devices, is implemented in SplitKey® technology, which is used in a popular signing solution Smart-ID®. Unfortunately, a two-party scheme that could create standards-compliant quantum-resistant signatures does not exist. This thesis presents a novel two-party signing scheme capable of creating ML-DSA-compliant signatures — Duolithium. This scheme is resistant against potential active attacks by either party, both during the key generation and signing processes. The thesis proposes some parts of Duolithium that were invented as a part of this thesis research and documents the remaining parts with reliance on prior research. Additionally, this thesis presents a complete, tested for functionality implementation of Duolithium in Python, together with the results of the benchmarks of network overhead and computational performance. The benchmark results suggest that Duolithium may be used to implement a new, quantum-resistant version of SplitKey that would be fully compatible with any signature verification component that supports ML-DSA.
  • Kirje
    Collaborative Multi-Agent Architecture for Domain-Agnostic Named Entity Recognition
    (Tartu Ülikool, 2025) Nabiyev, Rasul; Šuvalov, Hendrik, juhendaja; Masing, Karl-Oskar, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Named Entity Recognition(NER) traditionally requires extensive domain-specific training data to achieve satisfactory performance for a given domain. Recent advancements in large language models have enabled the development of NER systems without supervised training, though this approach still requires careful prompt engineering and may need external knowledge augmentation during inference. This thesis introduces a novel domain-agnostic NER framework based on a collaborative multi-agent architecture that can adapt to any domain given only entity definitions and their descriptions. The framework consists of 4 high-level components: a team of agents, a metaprompter, a chat supervisor and a grounding engine. The system requires no training data or prompt engineering for new domains, operating as a few-shot solution for NER tasks. The framework's performance is evaluated across 4 distinct domains using standard NER benchmark datasets. Our evaluation shows that the multi-agent approach outperforms the baseline of few-shot NER with single LLM call in 3 out of 4 benchmarks, suggesting a promising direction for domain-agnostic NER. Ablation studies demonstrate varying effectiveness of each component on the system's performance depending on the domain, with the combination of three specialized agents and grounding engine proving generally most effective in all tested domains.
  • Kirje
    A Recommender System for Improved Data Findability in Open Government Data Portals
    (Tartu Ülikool, 2025) Huseynov, Ramil; Nikiforova, Anastasija, juhendaja; Symeonidis, Dimitrios, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Despite the large amount of data available through OGD (Open Government Data) portals, most of it remains “dark data” which means it is not being used. A significant factor contributing to it can be the usability challenges such poor data findability and discoverability associated with these portals. One of the ways to contribute to the solution of these challenges is a recommendation system that can suggest related datasets. Unlike other domains, the recommendation system in the OGD portals is special as it can’t rely on user profile as most OGD portals don’t require authentication. Moreover, this recommendation method should be adaptable to the diverse structures of these portals. Finally, existing recommendation systems for OGD portals mostly focus on tags/category recommendations not datasets recommendations or fail to capture the semantic meaning of dataset’s metadata when making recommendations. This study focuses on these challenges by proposing a new datasets recommendation method based on dataset’s metadata that can capture its semantic meaning without relying on user’s profile and compatible with wider range of OGD portals. To capture the semantic relations between dataset’s metadata the proposed recommendation system relies on pretrained Word2Vec model. Additionally, the prototype of the proposed recommendation system was implemented for the usability testing and feedback was collected and analyzed.
  • Kirje
    Developing a Human-Centric Training Method to Educate High School Students on Social Engineering Techniques
    (Tartu Ülikool, 2025) Pohjaranta, Santeri Rikhard Artturi; Mubashar, Iqbal, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Data security plays a vital role in our society. We use different tools for communication, such as social networks, emails, and phone messages. In the security of personal data, an important role play is technology, which collects and secures user data and the human who owns it. If technology helps to secure data, then the human role in the system is to hold access to their data and not give it to another person. From different papers, it could be found that the “user is the weakest link in the security chain”. This happens because of various psychological manipulations that attackers use to receive sensitive data from users or, in other words, Social Engineering. To prevent such situations, people must be taught how attackers could receive their data through such manipulations and how to not fall into an attacker's trap by creating human-centric cybersecurity training. The current solutions lack a human-centered approach and platform tailored to high school students. Therefore, this research provides information about weak social engineering spots among high school students. Using knowledge about high school students' weak social engineering skills, this research presents a game-based training program using a one-platform solution to train high school students against current social engineering techniques with which they have problems. The efficiency of the training and platform is evaluated by the results of the first and second questionnaires to provide results of changes in the social engineering knowledge and skills of high school students.
  • Kirje
    Bird Colony: an Application for Collecting Spatial and Fitness Data from Seabird Colonies for Analysing Global Patterns
    (Tartu Ülikool, 2025) Meitern, Richard; Sepp, Tuul, juhendaja; Vasser, Madis, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Seabirds are vital indicators of environmental change due to their wide-ranging habitats and ecological sensitivity. Hence effective monitoring of seabirds is crucial for global understanding of environmental change. To better utilize the ongoing research efforts that collect data about seabird breeding there is a need for software tools that facilitate data collection in seabird research. This thesis introduces the Bird Colony application, a user-friendly tool developed with Flutter and Firebase to streamline data collection and management in seabird colony monitoring. Key features of the developed application include precise nest mapping, real-time data entry, and experiment management - all tailored to the specific needs of field researchers. The application's cross-platform compatibility and real-time database enable immediate data synchronization between researchers working in a colony. The open-source nature of the application promotes customization and collaboration within the developer community, potentially improving seabird research methodologies and conservation strategies in the long run. To test and validate the usefulness of the developed software the Bird Colony application has been effectively used on Kakrarahu islet, Estonia to improve the monitoring of its common gull (Larus canus) breeding colony. On Kakrarahu common gulls have been monitored for several decades — all in an effort to enhance the understanding of seabird populations and their responses to environmental change. By improving data collection methods and fostering collaboration among re-searchers, this application aims to contribute significantly to seabird conservation and ecological research worldwide.
  • Kirje
    Formidable Fortress – A Level Based Artillery Game Featuring Rule Based Dynamic Difficulty Adjustment
    (Tartu Ülikool, 2025) Islam, Euna; Muhhin, Mark, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    This thesis presents the design and development of Formidable Fortress, an artillery game featuring a rule-based Dynamic Difficulty Adjustment (DDA) system. Traditionally, games offer static difficulty settings at the beginning of the game. This method does not consider that the player’s skill may improve during the game. DDA dynamically adjusts the game's difficulty to match the player’s skill level and performance. This study explores a rule-based approach to DDA. The system has an initial evaluation phase where players from all experience ranges face the same level. After this evaluation, the game identifies the player’s level and starts acting accordingly. The game was developed using the Unity game engine, and art assets were created in Aseprite. Playtesting sessions were conducted with interviews to evaluate the effectiveness of the rule-based DDA system. Qualitative feedback and observational data were collected to check player performance and engagement and identify design issues. The results demonstrate that the system mostly successfully adjusts difficulty to individual player skills after the initial evaluation phase. Limitations were identified from the player feedback, where experienced players found the initial skill evaluation phase overly simple. The study concludes that a rule-based DDA approach provides a practical solution for adaptive gameplay. This system may work in small-scale projects or games with limited computational resources. Future research should explore how to determine players' skill levels more efficiently and how to reduce the designer’s workload.
  • Kirje
    Developing a Testing Framework for Internet of Things Systems using IoTempower as an Example
    (Tartu Ülikool, 2024) Heydarov, Araz; Norbisrath, Ulrich, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    The Internet of Things (IoT) has made technology much more powerful, enabling smart features to be built into every aspect of daily life. However, this involves a lot more than developing a device that connects to the internet – IoT development is complicated and difficult. This can be especially daunting for beginners and even industry professionals as well. One challenge faced by developers, rather than users, is building a reliable IoT framework, as testing is not a common practice across these platforms. IoTempower is no exception. IoTempower was created as an accessible framework for engaging with IoT that could be used by anyone from tinkerers and programmers all the way up to students, artists, or professionals–and everyone in between! It is also great for schools because not only does it teach about home automation systems with real-world applications, but it also serves as a powerful teaching tool for higher-level concepts surrounding internet-enabled devices. This thesis focuses on regression testing and hardware management within the IoTempower framework, with the broader aim of studying how to effectively test hardware frameworks. The main aim is to develop a complete test suite that allows new features to integrate seamlessly without breaking existing ones while adding new hardware support. Enhancing these areas provides a better user experience, makes the framework more reliable overall, and advances testing methods adopted in various IoT development frameworks, thus increasing reliability in different IoT scenarios.
  • Kirje
    Python Programming Module for Non-specialized Schools as an Introduction to IT
    (Tartu Ülikool, 2024) Dorokhov, Mykhailo; Lepp, Marina, juhendaja; Bardone, Emanuele, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    This thesis describes the motivation behind, the emergence, and the subsequent evolution of the IT module, a set of three courses designed by the author, that has been running as an elective module in Tartu Annelinna Gümnaasium since 2020. The module combines Python programming language learning and project-based teamwork, serves as an introduction to the IT industry and provides career guidance for school students by presenting them the possibilities to study Computer Science at the University of Tartu. The aim of the thesis is to present a motivation that led to the creation of a new elective module in the school, describe how it’s integrated into the state and the school’s curriculum, break down course structure, explore the discovered teaching techniques, and reflect on the continuous improvements that the 4 iterations of the module have experienced. As a result, the best-suiting digital learning environment setup (Discord, Replit, Notion), a set of effective teaching techniques (pop-culture references, analogies, real-life-inspired homework and incremental learning) and events (guest lessons, “Chasing Unicorns” movie session, the University of Tartu Delta center visit and Pipedrive office visit with pizza party for the high-achievers), as well as a well-structured sequence of topics, has been established and described. The module’s mate-rials are shared via a public link for fellow teachers and researchers.
  • Kirje
    Model Drift in Federated Learning: an Experimental Analysis
    (Tartu Ülikool, 2024) Rahimli, Leyla; Awaysheh, Feras Mahmoud Naji, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Our ability to extract knowledge beyond data silos will drive the future of machine learning towards more accurate and comprehensive models. Federated Learning (FL) has emerged as a transformative paradigm in machine learning, enabling decentralized model training while preserving data privacy across multiple clients. By distributing the learning process, FL addresses critical privacy concerns but introduces challenges related to model drift. Model drift is the phenomenon where a model degrades over time due to changes in the underlying data distribution or the relationships between input features and target variables. This issue is especially pronounced in FL environments, where data is not independently and identically distributed (non-IID) across clients, leading to asynchronous and heterogeneous updates that intensify drift. In response to the challenge of model drift in FL, this thesis proposes a novel methodology for drift detection and management within federated environments. By implementing the Flower federated learning framework integrated with Alibi Detect, a specialized tool for drift detection, the study introduces an effective strategy to monitor and identify both concept drift (changes in the relationship between inputs and outputs) and data drift (changes in the input data distribution). The proposed methodology uses statistical tests to accurately detect significant deviations in model performance, ensuring timely intervention and model updates. Our experimental analysis demonstrates the effectiveness of the proposed drift detection framework. By simulating FL scenarios with varying degrees of drift introduced across different clients, the study systematically evaluates the impact of drift on model performance metrics, including accuracy, F1 score, Cohen's kappa, and ROC. The findings indicate that even minimal drift in a subset of clients can significantly degrade the global model's performance, underscoring the importance of robust drift detection. The proposed solution enhances the reliability and accuracy of federated models and addresses the scalability and privacy-preserving requirements inherent in FL environments. The contributions of this thesis are significant for the future development and application of FL systems. This study paves the way for more resilient FL models capable of maintaining high performance in dynamic and distributed settings by providing a framework for detecting model drift. The implications of this work extend to various domains where FL is employed, such as healthcare, finance, and personalized services, where the accuracy and reliability of models are critical. This research sets the foundation for future explorations into more advanced drift management techniques, ultimately contributing to FL's broader adoption and efficacy in real-world applications.
  • Kirje
    Digital Twin Technology in Financial Services
    (Tartu Ülikool, 2024) Kazakbaeva, Azhar; Milani, Fredrik Payman, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    The focus of this thesis is the exploration of potential use cases of digital twins in financial services. This study explores the use cases, benefits, and challenges of the implementation to define the value and impact this technology can bring to financial services. As a result of the case study, four potential use cases of digital twins were identified, as well as four benefits and four challenges of digital twin adoption. The main contribution of this research is a framework that can help practitioners and researchers understand how digital twins can be utilized in financial companies
  • Kirje
    Analysis of eTwinning Projects Related to Informatics
    (Tartu Ülikool, 2024) Bork, Liina; Luik, Piret, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    The purpose of this master’s thesis is to determine how informatics is implemented in Estonian schools through eTwinning projects in the year 2022-2023. During this process, projects in the eTwinning TwinSpace environment were studied, with the final sample consisting 37 projects. The collected data included general information about the projects, the use of digital competence and informatics competence in informatics projects and the use of information and communication technology tools and resources during the projects, categorized according to their main functions. Data analysis was conducted using descriptive statistics and quantitative methods, with Microsoft Excel used to support the interpretation of results and drawing of conclusions. The research showed that eTwinning projects involving informatics were carried out in several counties and various educational institutions in Estonia, with a primary focus on developing digital competencies. Although a variety of digital tools were used in the projects, the integration of computer science competencies was modest, particulary in schools where computer science was not part of the curriculum. The author of the thesis recommends that future efforts focus more on competencies of informatics and that additional training be provided to teaches to improve the integration of computer science into eTwinning projects.
  • Kirje
    Analysis of Bicycle Sharing Data for Decision Support to Expanding Tartu Cycling Infrastructure
    (Tartu Ülikool, 2024) Tarro, Martti; Pourmoradnasseri, Mozhgan, juhendaja; Tera, Helen, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Public bicycle sharing systems have seen an increase in use across many cities in the world, and their expansion is partially dependent on the state of the infrastructure. Traditional road planning procedures rely on empirical evidence and surveys. The widespread availability of GPS data from micromobility sharing systems have seen the approaches enhanced by the analysis of movement patterns. Tartu has strategic plans for expansion of its cycling network to make the city as well as its surroundings more accessible by cycling. This thesis examines the patterns of Tartu Bike Share users using their geolocation data, and compares the planned networks to proposed paths from four strategies for prioritisation of new road sections. The strategies focus on evaluating current cycling patterns, estimating optimal paths, finding ways to make the current network more cohesive, and a combination of these strategies through MULTIMOORA modelling. The prioritised gaps from the model include multiple potential cycling paths, that had not been included in the planning of the main and auxiliary networks. The cycling network in Tartu is already expansive, but identifying significant gaps in the current and planned networks has the chance to improve it yet more.
  • Kirje
    Reviewing the Classification Performance of Recent Neuro-fuzzy Systems
    (Tartu Ülikool, 2024) Soni, Ayushmat Bhardwaj; Tomasiello, Stefania, juhendaja; Uzair, Muhammad , juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    This thesis explores the classification performance of recent neuro-fuzzy systems, which combine fuzzy logic with neural networks to enhance machine learning applications. Through a meta-analysis of literature from the past decades, the study evaluates various neuro-fuzzy architectures including the classical Adaptive Neuro-Fuzzy Inference System (ANFIS) and their performance in different domains. Comparative analyses with different approaches highlight neuro-fuzzy systems' strengths in handling imprecise and noisy data and their dependency on fuzzy set design and neural architecture. The goal of this work is to offer practical insights for both practitioners and scholars on the selection and management of appropriate methods for classification tasks across various application domains such as medicine and finance. This is achieved through a detailed analysis of diverse approaches, addressing a gap in recent comprehensive reviews of these methods.
  • Kirje
    Towards a Novel Taxonomy for Requirements Interdependencies
    (Tartu Ülikool, 2024) Mirzazada, Elvin; Gharib, Mohamad, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Requirements interdependencies capture how requirements relate to and affect each other, and it is one of the main problems to be handled for delivering quality requirements, and in turn, software systems with high quality. This said, requirements interdependencies cannot be overlooked or ignored and must be properly handled since requirements dependencies influence several design, development, and implementation decisions, and inappropriate handling of such interdependencies can lead to software system development failures. Although various interdependencies among requirements have been considered in the literature (e.g., requires, refines, similar, or conflicts), they are not able to cope with the advancement on the requirements side. More specifically, systems are becoming more complex, leading to more complex interdependencies among their requirements, which available interdependencies might not be able to capture. This thesis aims to solve this problem by developing a new taxonomy of requirements interdependencies that can better understand software requirements and the dependencies between them. The taxonomy aims to overcome the limitations in existing work by proposing a taxonomy that offers a comprehensive set of requirements interdependencies. To achieve that the taxonomy has been mined via a Systematic Literature Review (SLR). The new taxonomy aims to facilitate the production of a more elaborated and expressive set of software system requirements, which will positively contribute to the development of high-quality software systems. As a result, this thesis aims to provide a solution to a problem encountered in the field of software development, to make requirements analysis more effective and efficient, and to contribute to the production of higher-quality software. This thesis examined the relationship between software requirements and dependencies in-depth and identified 16 different types of relationships. These relationships are classified into different categories. This thesis offers suggestions for future research to address issues such as expanding the application areas of the taxonomy, including new dependency types and developing automation tools.
  • Kirje
    Public Perception of Artificial Intelligence in Ukraine and Russia: News Media and Social Media
    (Tartu Ülikool, 2024) Chemodanova, Olga; Sharma, Rajesh, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Artificial intelligence is essential in various fields in today’s society. Certain AI advancements have gained popularity and are being actively utilized by the public. In order to study how AI affects society and to establish appropriate regulations to guide its development, one can examine people’s opinions about AI. The public’s views incorporate expectations and emotions regarding the technology, especially its potential negative impacts. This analysis can assist in implementing ways to adjust society to the advances of AI. While AI is becoming more prominent, public perceptions, especially in Eastern Europe, have not been thoroughly investigated, this research closes the gap by examining the views of Ukrainian and Russian people on AI by analyzing online news media and social media discourse. Here we demonstrate that news mostly highlights the innovations in AI and its advantages, often neglecting potential hazards, whereas social media users frequently engage in discussions about a wider array of subjects, encompassing potential risks of AI. These findings indicate that in Ukraine and Russia, news mainly pushes the narrative of the advancements of AI to the general public, which can lead to the adoption of a one-sided perception of the technology by people. However, there is an interest in discussing possible risks of AI, reflected in social media users’ AI anxieties. Grasping these viewpoints is essential for creating a thorough and balanced discussion of AI and its conseques.
  • Kirje
    Daily Real-time System for Assessing Urban Traffic Emissions Equilibrium Using IoT Data
    (Tartu Ülikool, 2024) Biswas, Ritudisha; Hadachi, Amnir, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    Global warming, driven by escalating greenhouse gas (GHG) emissions, is one of the most critical environmental issues facing the world today. One promising strategy for mitigating global warming and its associated impacts is the integration of green spaces within cities. This thesis aims to explore the critical relationship between streetlevel emissions and the presence of green areas in urban environments. By monitoring emissions at the street level and assessing the carbon sequestration potential of urban green spaces, we seek to understand how effectively green spaces can mitigate the impacts of urban pollution. The thesis underscores the necessity for real-time monitoring systems that provide actionable insights for urban planners and policymakers to optimize green infrastructure and reduce urban pollution impacts. The findings emphasize the role of urban greening in creating healthier cities better equipped to handle environmental challenges.
  • Kirje
    Comparison of National Exam Tasks and ChatGPT Rephrased Tasks
    (Tartu Ülikool, 2024) Keivabu, Carmen; Luik, Piret, juhendaja; Orav-Puurand, Kerli, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    The goal of the master's thesis was to determine to what extent mathematics word problems rephrased using ChatGPT help simplify the understanding and solving of complex mathematical problems for students and improve their academic performance. The study was conducted among 26 11th-grade students from two high schools in Lääne-Virumaa. The students solved four national exam problems as well as their rephrased versions by ChatGPT. A mixed research method was used, including both quantitative and qualitative data collection. The students' results were analyzed using statistical methods in the JASP program, and feedback on the clarity and comprehensibility of the tasks was gathered through a Google Form survey. The main findings showed that students generally performed better on the problems rephrased by ChatGPT compared to the national exam problems. Based on the feedback, some tasks were considered clearer in the national exam versions, while others were more understandable in the versions rephrased by ChatGPT. In conclusion, ChatGPT has the potential to support mathematics teaching, but its effective use requires careful formulation of tasks.