MTAT magistritööd – Master's theses

Permanent URI for this collectionhttps://hdl.handle.net/10062/30974

Browse

Now showing 1 - 20 of 1112

3D graafi uurimine
(Tartu Ülikool, 2010) Danilov, Dmitri; Norbisrath, Ulrich; Tartu Ülikool. Matemaatika-informaatikateaduskond; Tartu Ülikool. Arvutiteaduse instituut
Ühiskonna üheks suureks probleemiks on informatsiooni üleküllus ning sellest tulenevalt on otsingutehnoloogia muutunud järjest olulisemaks. Teadlasi motiveerib arendama aina intelligentsemaid otsingusüsteeme see, et informatsioon on tihti korduv ja vajaliku informatsiooni leidmine keeruline. Kui otsingu ülesanded sisaldavad rohkem kui üht probleemi, siis on oluline arendada süsteem, mis suudaks koondada informatsiooni ja lihtsustada kogu otsingu protsessi. Üks paljulubavatest uuringuvaldkondadest on exploratory search (edaspidi uuriv otsing). Selle eesmärgiks on aidata otsijal õppida ja uurida uurimisprotsessi käigus. Antud magistritöö eesmärgiks on realiseerida uuriva otsingu visuaalne programm. Valminud programm visualiseerib graafi ruumiliselt. Navigatsiooni ja kasutajaliidese lahendused võimaldavad uurida graafi andmeid ja luua hüpoteese, mis baseeruvad leitud andmetel ja andme struktuuridel. Programm joonistab graafi kasutades 2.5D paigutust. Graaf on struktureeritud ja koosneb ringjatest tasapinnalistest alamgraafidest. Antud rakendus loob automaatselt klastreid (cluster) joonistamise käigus. Autori poolt loodud programmi juba kasutatakse mõndades projektides ja seda on lihtne laiendada teistele graafi andmetele. Autor leiab, et antud projekt on suure potentsiaaliga ning tulevikus võib seda programmi effektiivselt kasutada keeruliste andmekogumite uurimiseks.
3D graph exploration
(Tartu Ülikool, 2010) Danilov, Dmitri; Norbisrath, Ulrich, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
3D-prinditava kere disain ja analüüs vabavaralisele haridusrobotile Robotont
(Tartu Ülikool, 2023) Mõtshärg, Eva; Vunder, Veiko, juhendaja; Raudmäe, Renno, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Haridusrobootika maailmas on palju erinevaid roboteid, kuid napib lahendusi, mis sobiksid üheaegselt nii algajale kui edasijõudnud õppurile. Tartu Ülikoolis arendatud Robotont on loodud olema sild kooli- ja tööstusrobootika vahel. Siiski on ka Robotondiga keeruline õpetada paljusid algajatele vajalikke kontseptsioone, kuna platvormi kasutusvõimalused on piiratud ning see ei ole paindlikult muudetav. Samal ajal kogub 3D-printimine järjest populaarsust ja pakub alternatiivi traditsioonilistele tootmismeetoditele. Käesolev magistritöö uurib ja analüüsib Robotondi 3D-printimise otstarbekust. Töö metoodikas rakendatakse välearenduse põhimõtteid, arvutijuhitavaid mõõtmisi ja semantilist versioonihaldust. Töös on lisaks valdkonna teoreetilisele ülevaatele kirjeldatud Robotondi 3D-prinditava kere disainiprotsessi ja põhjendatud tehtud valikuid. Empiiriliste katsete toel töötati välja 3D-printimise protsess, mis tagab Robotondi õppetöös kasutamiseks piisava vastupidavuse. Kasutades töö raames loodud 3D-mudeleid ja optimeeritud G-koodi faile on igaühel võimalik 3D-printida endale oma Robotont ja hõlpsalt kohandada see oma vajadustele vastavaks.
A Blockchain Solution for Auditing of Timber-to-Charcoal Process
(Tartu Ülikool, 2021) Zubair, Muhammad; Iqbal, Mubashar, juhendaja; Milani, Fredrik Payman, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Blockchain technology is emerging not only in cryptocurrencies like bitcoin and ethereum but also expanding in other business fields. Due to its diverse nature of architecture, it can easily integrate with such environments, e.g., permissioned or permissionless, state-full or stateless. This article mainly focuses on permissioned blockchain, the hyperledger fabric, how to implement it, and how to use it to automate the audit to the timber-tocharcoal business process. In order to accomplish this study, a design science approach is selected. In this thesis, a timber-to-charcoal case study is analyzed, designed the software architecture, developed the solution using hyperledger fabric, tested and evaluated it with defined evaluation criteria. The technology stack to develop the smart contract and web services to invoke the chaincode and access the ledger is NodeJs. The final evaluation shows that hyperledger fabric is a secure, fast, and decentralized immutable ledger. In conclusion, it is a suitable approach and benefits the audit to timber-to-charcoal business process.
A Case Study on Post-editing Machine Translation: Tasks, Challenges, and Attitudes
(Tartu Ülikool, 2021) Shuyler, Katrin; Nolte, Alexander, juhendaja; Fišel, Mark, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
While neural machine translation is gaining wider commercial traction as a useful tool for translating technical texts, the professional translator community is hesitant about post-editing machine translation. A qualitative case study guided by the socio-technical idea of the importance to achieve a balance between human and technical aspects of a system was conducted with a focus on post-editing machine translation. The eight translators who took part in this study were given an identical task of post-editing machine-translated technical text. Interviews were then conducted allowing translators to express their perspectives. The interview questions were designed to address how translators approached the study task, what caused them difficulties about post-editing, and their attitudes towards machine translation. The analysis of the data collected for this case showed that while there was variation to how the translators approached the post-editing task, there appeared to be a workflow shared by the majority of participants. Regarding the challenges translators faced in post-editing, the analysis suggests the key factors affecting the translators were decision making and knowledge of the translation tool. This research contributes to the knowledge about translators’ attitudes by showing a difference of opinion between professional and personal machine translation use.
A Collaborative Approach for Large-scale Electricity Consumption Using Federated Learning
(Tartu Ülikool, 2022) Özen, Canberk; Awaysheh, Feras, juhendaja; Alawadi, Sadi, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Forecasting energy demand is a crucial topic in the energy industry to keep the balance between supply and demand, hence keeping the grid in effective operation. The adoption of renewable energy sources for the supply makes the forecasting problem ever the more prominent because of the additional uncertainty they bring to the grid, besides the consumers’ energy usage patterns. The uncertainty on the demand side forecasting can be theoretically overcome via a centralized predictive model that takes note of the consumers’ past electricity usage. However, in practice, forecasting energy demand is challenged by users’ concerns for the privacy of their energy data and the scalability of storing it, in addition to completing the model updates in time. Both problems can be solved if the centralized training paradigm is replaced with federated training, where each household trains its model locally, and the centralized server only acts as a coordinator by aggregating the weights of the individual models’ and sending the updates back to them, all without seeing the consumers’ data. Because of the diversity in energy usage, the convergence of local models may require too much time. This study will investigate federated learning to develop a clustering algorithm that groups similar residences as one node to fasten the model convergence without reducing its accuracy.
A Comparative Evaluation of Explainability Techniques for Image Data
(Tartu Ülikool, 2024) Skliarov, Mykyta; ElShawi, Radwa, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Masinõppe kasutamine on viimasel kümnendil paljudes valdkondades märkimisväärselt kasvanud, eriti arvutinägemise valdkonnas, kus suure jõudlusega konvolutsioonilised süva-neuronivõrgud on saavutanud ja isegi ületanud inimese jõudluse paljudes valdkondades. Et vastata nende black-box mudelite läbipaistvuse kasvavale vajadusele, on eXplainable AI kogukond loonud erinevaid tehnikaid nende prognooside selgitamiseks. Populaarne viis seda teha pildiandmete puhul on kasutada silmapaistvuse kaardid. Samas ei ole nende meetodite kvaliteedi objektiivne hindamine tõlgendatavuse mitmetahulisuse tõttu lihtne ülesanne. Käesolevas töös teostame kuue populaarse silmapaistvuse kaardi selgitamise tehnika, nimelt LIME, SHAP, GradCAM, GradCAM++, IntGrad ja SmoothGrad, põhjaliku võrdleva hindamise, kasutades kirjanduses esinevaid viit kvantitatiivset funktsioonipõhist mõõdikut, täpsemalt truudust, stabiilsust, identiteeti, eraldatavust ja aega, kolme üldkasutatava võrdlusandmestiku ja kolme tuntud mudelite arhitektuuri põhjal, et määrata kindlaks iga tehnika plussid ja miinused. Kuigi me leiame, et ükski tehnika ei domineeri kõigis mõõdikutes, näitavad saadud tulemused, et IntGrad ja SmoothGrad esinesid meie usaldusväärsuse ja stabiilsuse testides hästi, kusjuures SHAP saavutas ka kõrgeid tulemusi usaldusväärsuse osas. Kõik meetodid peale LIME ja SmoothGrad said kõrgeid tulemusi identiteedi mõõtkavas ja kõik peale LIME - eraldatavuse mõõtkavas, samas kui GradCAM ja GradCAM++ olid kaugelt kõige kiiremad. Samuti märgime, et meetrikate puhul on ettevaatusabinõusid, mis viitavad sellele, et erinevate XAI-tehnikate kvaliteedist täieliku pildi saamiseks on vaja teha rohkem tööd.
A comparative study of database schema on query performance in data warehousing
(Tartu Ülikool, 2024) Pajuleht, Roland; Ševtšenko, Eduard, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Käesolevas töös võrreldakse päringute jõudluse ja täitmisplaanide erinevusi kahe erineva andmemodelleerimise lähenemisviisi vahel. Mõõtmelist modelleerimist, mis on andmelao ehitamise üheks standardlahenduseks, võrreldakse vähem standardiseeritud mudeliga, mis hakkab tekkima, kui konkreetsed andmearhitektuuriprotseduurid ei ole paigas. Nende skeemide peal viiakse läbi mitmed analüütilised. Selgus, et kuigi loetavus laia tabeli puhul on parem, tekivad kiiresti jõudlusprobleemid. Traditsioonilistes andmeladudes töötades peavad andmetöötlejad järgima väljakujunenud arhitektuuripraktikaid, et hoida üleval tõhusat andmeladu.
A Comparison of Privacy Enhancing Technologies in Internet of Vehicle Systems
(Tartu Ülikool, 2020) Onyinye, Nwaokolo Anita; Matulevicius, Raimundas, juhendaja; Affia, Abasi-amefon Obot, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
The evolution of internet of things (IoT) is driving conventional vehicle ad-hoc networks into the internet of vehicles (IoV). IoV has shown great relevance in helping to improve traffic efficiency by easing traffic congestion. Vehicles can easily exchange information with other vehicles and infrastructure in the same environment as they are. However, the exchange of information in IoV introduces privacy challenges to vehicle owners. Private data of these IoV users are being leaked unintentionally to the system, some systems even send user’s data to a third party system and most times the user’s are not aware of such message exchange with their private data. It is possible for a honest but curious IoV user to identify vehicles and track user’s location by analysing messages exchanged between systems. Messages exchanged between these systems carry some form of vehicular identification and in turn can be used to trace an owner via location profiling. This thesis follows a structured approach towards mitigating this privacy leakage by applying the privacy enhancing technologies (PETs) identified in this thesis paper (encryption and attribute based credential), to help in protecting the information exchanged between each IoV’s involved in communication. This approach is beneficial to system entities compliance in privacy frameworks and also helps in identifying the most effective PET by comparing the state and condition of each data objects identified in the system.
A Competitive Scenario Forecaster using XGBoost and Gaussian Copula
(Tartu Ülikool, 2023) Kolomiiets, Denys; Shahroudi, Novin, juhendaja; Kull, Meelis, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
In recent years scenario forecasting has been explored and developed by multiple authors. It is a useful technique for setting such as renewable energy production, which is extremely important for a society transitioning from fossil fuel energy generation. Currently, one of the methods to approach the task of scenario forecasting are generative models. The primary goal of this thesis is to develop an approach that outperforms the current best model, using the decision tree model method. This work also discusses possible improvements for decision tree models in scenario forecasting setting. Our approach has surpassed the performance of generative models, making it a solid new baseline for future researchers to beat.
A conjugate gradient solver library for the Playstation 3
(Tartu Ülikool, 2010) Laasik, Toomas; Norbisrath, Ulrich, juhendaja; Vainikko, Eero, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
A conjugate gradient solver on the PlayStation 3 - a native approach
(Tartu Ülikool, 2010) Tulmin, Lauri; Norbisrath, Ulrich, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
A dashboard to visualize the product quality level
(Tartu Ülikool, 2021) Prink, Kert; Scott, Ezequiel, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Software releases nowadays happen faster than ever. People responsible for the software products are called Product Owners (POs) in a modern popular framework called Scrum. The role of the PO is difficult because it requires making decisions based on the business and technical data. Successful POs need an overview of all the variables to make decisions with the best value. Software quality is one such variable that the POs must consider while making decisions. There are tools that provide data about the software quality, but there is a lack of tools that give a straightforward overview of the quality in a way that is easy to understand. In this thesis, an open-source software application was delivered that provides a straightforward overview of the quality level on each release. The application was designed so that additional future developments could be easily made. Firstly, the requirements for the application were elicited. Secondly, the application was implemented. Next, the application was tested to verify functional correctness. And finally, the application was evaluated with a survey study along the POs in the industry to understand the perceived usefulness and ease of use of the application.
A Decentralized Public Key Infrastructure for Trust Management in X-Road
(Tartu Ülikool, 2023) Leung, Kin Long; Bakhtina, Mariia, juhendaja; Awad, Ahmed, juhendaja; Matulevičius, Raimundas, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Today, Public Key Infrastructure with X.509 (PKIX) is the building block for establishing secure connections over the Internet and creating digital signatures. In PKIX, Certificate Authority (CA) is responsible for the creation of certificates and the resolution of certificate statuses. Due to the centralized architecture, CA becomes a single-point-offailure to any network that relies on it to establish trust. By utilizing distributed ledger technology (DLT), decentralized identifiers and verifiable credentials can be verified without intermediates like CAs. They can be used to construct a Decentralized Public Key Infrastructure (DPKI) which eliminates the shortcomings of PKIX. In this thesis, we studied X-Road, a centrally managed distributed data exchange system depending on PKIX, and presented an alternate DPKI architecture that uses DLT-based decentralized identifiers and verifiable credentials to build up trust between information systems. A proof-of-concept was implemented and evaluated. The findings demonstrate that the alternative DPKI architecture enhances the trustworthiness of the data exchange system, particularly in terms of security and reliability.
A Desktop Application for Advanced Business Rule Mining
(Tartu Ülikool, 2020) Alman, Anti; Maggi, Fabrizio Maria, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Process mining is one of the research disciplines belonging to the field of Business Process Management (BPM). The central idea of process mining is to use real process execution logs in order to discover, model, and improve business processes. There are multiple approaches to modeling processes with the most prevalent being procedural models. However, procedural models can be difficult to use in cases where the process is less structured and has a high number of different branches and exceptions. In these cases, it may be better to use declarative models, because declarative models do not aim to model the end-to-end processes step by step, but they constrain the behaviour of the process using rules thus allowing for more variability in the process model. There are multiple applications available for working with procedural models. For example, Disco and Apromore, both of which have a highly polished user interface and are relatively easy to use. However, there are currently no comparable applications for working with declarative models. This thesis builds on the Master’s Thesis of D. Kapisiz in order to develop an already existing application, RuM, into an accessible and easy to use process mining application. While RuM itself already has most of the needed functionality, the user interface of RuM is not well polished and does not have an appealing look in general. In this Master’s Thesis we will completely redesign and reimplement the user interface of RuM while also making technical changes in order to enable its continued development. The new user interface has been thoroughly evaluated by conducting a user evaluation involving 4 experts of declarative models and 4 experts of business process mining in general. The main findings of the user evaluation will be presented as a part of this thesis.
A Functional Prototype and General Architecture of An-alytic Data Management for a Railway Company
(Tartu Ülikool, 2024) Metelitsa, Mait; Raun, Kristo, juhendaja; Awad, Ahmed, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Magistritöö käsitleb andmetehnika ja ärianalüütika temaatikat raudtee-ettevõttes. Töö ees-märk on lahendada EVR-i analüütilise andmehaldusplatvormi moderniseerimisega seotud väljakutseid. Lõputöö üks peamisi uudseid külgi on detailne analüüsipõhine kirjeldust sellest, kuidas EVR-il oleks kõige otstarbekam nii rahaliselt kui ka arhitektuuriliselt oma analüütilist and-mehaldusplatvormi moderniseerida. Töös on avatud hetkeolukord, kirjeldatud lähteandmete nomenklatuuri ning läbi otsustusdimensioonide ja tehnoloogiliste alternatiivide üksikasja-liku analüüsi on jõutud tulevase arhitektuuri teekaardini. Väljapakutud tulevikuarhitektuur toetab paindlikku hübriidset lähenemisviisi, mis ühendab infrastruktuuri kui teenust (IaaS) ja tarkvara kui teenust (SaaS) käitumise mudeleid. Rõhu-tamist väärib, et tasakaal IaaS ja SaaS tarkvara käitamismudelite vahel on kohandatud vas-tavalt andmeallikate tüübile ja keerukusele. Praktilise panusena realiseeritakse töös tulevikuarhitektuuri komponentidele tuginedes kaks kaasust. Esmalt, ostuarvete XML-ide ridade ning teiseks, ülesõitude logiandmete automa-tiseeritud analüüsi- ja andmetöötlusvood. Praktiliste kaasuste jaoks arendatud lahendused baseeruvad mh tehnoloogiatel nagu Dagster andmetöötlusvoogude orkestreerimiseks, Post-gres andmete säilitamiseks ja Streamlit andmete visualiseerimiseks. Sellega annab lõputöö praktilise panuse EVRi andmetöötlusvõimekuse edasisse parandamisse ja valmistab ette tehnilist võimekust ja asutusesisest kompetentsi SaaS andmejärve platvormi kasutuselevõ-tuks. Eeldatavalt saab selleks platvormiks olema Microsoft Fabric. Lõputöö annab üldisema panuse andmehalduse ja andmetehnika valdkonda, pakkudes struk-tureeritud raamistikku analüütiliste lahenduste arhitektuuridisaini loomiseks. Võttes arvesse andmetehnika ja ärianalüüsi suhteliselt mõõdukat varasemat vaagimist raudteesektoris, toetab käesolev lõputöö märkimisväärselt selle teadmislünga ületamist.
A Java adapter for Friend-to-Friend computing
(Tartu Ülikool, 2011) Škarupelov, Vladimir; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
A Metrics-based Approach for Evaluating the Quality of Microservices-based Systems
(Tartu Ülikool, 2024) Abakumova, Viktoriia; Gharib, Mohamad, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Tarkvaraarhitektuuri arenedes mikroteenustepõhiste süsteemide (MPS) suunas jääb selliste süsteemide kvaliteedi hindamine keerukaks ülesandeks. See doktoritöö tutvustab mõõdikute põhinevat lähenemist MPS-i kvaliteedi hindamiseks. Peamine eesmärk on arendada hästi määratletud mõõdikute kogum, mis tabaks mikroteenuste põhilisi omadusi, nagu modulaarsus ja skaleeritavus, ning oleks kooskõlas tarkvara kvaliteedi tööstusstandarditega. Uuringumeetod hõlmab disainiteaduse lähenemist, viies läbi süstemaatilise kirjandusülevaate, et tuvastada olemasolevad MPS-i kujunduspõhimõtted, kvaliteediomadused ja mõõdikud tuvastatud kujunduspõhimõtete hindamiseks/analüüsimiseks. Uuring jätkub, määratledes ainulaadse mõõdikute komplekti, mis sobib MPS-ile (lisaks olemasolevatele mõõdikutele on kujundatud ka mitu täiendavat mõõdikut), mis kujundatakse hiljem rakendatud lähenemisviisi abil realistlikule juhtumile. Tulemused rõhutavad iga mõõdiku tähtsust süsteemi kvaliteedi põhjalikuks mõistmiseks, sillutades teed tulevastele uuringutele ja praktilisele rakendamisele tarkvaraarenduses. Traditsiooniliste tarkvara hindamise mõõdikute ja mikroteenuste spetsiifiliste nõudmiste vahelise lõhe ületamisega pakub see doktoritöö väärtuslikke teadmisi arendajatele ja arhitektidele, kes soovivad parandada oma mikroteenuste arhitektuuride operatiivset sooritust ja kvaliteeti.
A Microservice-based Platform for Software Technical Debt Analysis
(Tartu Ülikool, 2021) Mulianingtyas, Octanty; Norbisrath, Ulrich, juhendaja; Rossi, Bruno, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Technical debt (TD) is defined as developing a new feature by a poor code and design to fulfil a punctual release date that may provide shorterm advantage, but negatively impact the future. Reducing TD is crucial to improve the software quality and decrease the financial issue. However, TD is cumbersome to calculate since every TD tool provided uses numerous metrics and models. This study aims to provide analysis about technical debt alternative measurements by developing a microservice-based platform on a set of large open-source projects. The result of the analysis is expected can be used as a reference for the researchers to arrange a proper formula to calculate TD. The microservice platform was selected since it has several benefits compared to monolithic applications. The benefits comprised simple deployment, fewer environment dependencies, and maintainability. We implemented three TD identification methods in the microservice platform: Maintainability Index (MI), SIG Maintainability, and Sqale. Through 124 directory java projects run in the microservice platform, the statistical analysis results show no correlation among the three TD identification methods. The increasing value of the metrics of TotalOp, TotalOpr, CC, Duplication, UnitSize, and NumberParameters, WMC will decrease the ranking of result in each TD Identification method, and thus decrease the overall TD. The scalability of the microservice is seen by the performance of the microservice by the total time for a query executed in the API. We used the load testing in the postman to see the total time. The API used for testing employed the aforementioned All TD Identification method. From the experiment, the time for a query executed for calculating 5 directories is 1m 40.38 s and it will continuously improve until 34m 20.70 s for 124 directories. We also tested the microservice platform for running in seven iterations. However, the number of iteration is not correlated with the time for a query executed. The time for a query executed is variated in every iteration and not increasing continuously as the increment of the number of iterations. Based on these facts, the microservice performance is only influenced by the number of directories upload instead of the number of iterations for executing.
A Model for Detecting and Resolving Conflicts in Features Extracted from App User Reviews
(Tartu Ülikool, 2022) Enitilo, Solagbade Ayodele; Gambo, Ishaya Peni, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Many app developers use user feedback to improve their app’s quality. User feedback is a type of a review that originates from different user types and covers various parts of the app. This thesis employs in-depth review analysis to guide app developers in the requirement elicitation process to identify areas of an app where an upgrade is essential. However, reviews from various users might conflict based on their perspectives and feature interest in the app. This thesis addresses the problem associated with detecting and resolving conflicting user reviews by formulating a robust taxonomy of conflict types, causes, and effects. This is because establishing conflicts in user reviews has its own set of semantic and lexical concerns, such as identifying the conceptual overlap that exists between user reviews and decoding the semantic inference of a review, i.e., the "why, how, and what" of an app feature. To find the solution to the previously described conflict, 64, 964 app reviews were tested and evaluated from three different mobile app sources. As a result a semantic rule-based tool was developed that integrates knowledge representation and linguistic rules to detect conflicts in reviews gathered from app users. The result emphasised the holistic awareness of domain knowledge for effective conflict identification and resolution in app reviews. In addition, the findings suggested a methodical approach for applying conflict analysis to strengthen software design requirements. As a conclusion, considering the comprehensive and interpretable disposition of the conflict identification rules, the thesis presents a solution to resolve the conflicts identified in app reviews.

Browse

Browsing MTAT magistritööd – Master's theses by Title