Masterarbeit
Machine Learning based Content Analysis Component for ConTED
Completion
2024/08
Research Area
Students
Lev Kuznetsov
Advisers
Description
Nowadays, data sovereignty on the web is requested by citizens and political institutions. Therefore, several projects, like Solid or the Next Generation Internet from the EU are working on a decentralization of the web to enforce higher data sovereignty and data privacy. With the decentralization of the Web, there will be various web resources acquired during runtime. However, external data not controlled by third parties can be also malicious or harmful. Thus, external data should only be used if it is trustworthy enough to use. If so, the user experience is likely higher, but untrustworthy data usage could also lead to worse user experience. Even data from an already pre-approved sources must be constantly checked, as the source may provide untrustworthy data at any point in time. In order to make decentralized web applications aware of trust, several approaches for computing a trust metric for web resources have been proposed. One such example is the ConTED framework. One of the core units of this framework is a content-based analysis component, which must calculate content metrics that will be used for trust evaluation every time data is received.
The goal of this master's thesis is to investigate the current state of technology at the intersection of the fields of machine learning and linked data contentual analysis for trust evaluations in ConTED. In addition to the research part of this master's thesis, sufficient attention should also be paid to the possibility of integrating and refining existing solutions to be able to use them in the context of web applications. Based on the scientific work produced by the authors of ConTED, this thesis shall develop a solution that combines the research areas described above to derive evaluation metrics for linked data, which should then be used for the final calculation of the trust framework.
The master's thesis includes an implementation of the machine learning based content analysis component of the ConTED trust framework based on currently existing machine learning solutions. The functional parts of the solution should improve the individual trust metrics presented in the content analysis component of ConTED. The thesis should aim on evaluation and comparison of most suitable machine learning models for the aforementioned analysis in the use case of linked data. Additionally, the final solution is to be developed with focus on accuracy and runtime metrics. Finally, the results of the thesis should provide a comprehensive analysis of the solution.
The objective of this master's thesis is to find an approach or combination of approaches to solve the previously mentioned problem in the context of content analysis of linked data for trust awareness via ConTED. This particularly includes the state of the art regarding Trust Awareness and content-based analysis of linked data. The demonstration of feasibility with an implementation demonstrator of the concept is part of this thesis as well as a suitable evaluation with exemplary use cases.
