The exponential growth of digital platforms has led to a massive amount of user-generated content. This vast influx of mostly textual data provides great potential for insights, but also poses unique challenges for processing and management. Erjon Skënderi’s dissertation introduces novel techniques to address the issue by assessing traditional and neural network-based text representation methods.
“The integration of various text presentation methods through ensemble learning provides a robust and effective solution, greatly improving the performance in multiple classification applications and duplicate detection tasks. This approach reflects a model shift in how we analyse and manage Big Social Data,” Skënderi says.
The research contributes to current societal interests by enhancing the way online platforms understand and manage user-generated content. It provides practical applications that could have a positive impact on query database management, document categorisation, and the real-time analysis of social trends. The methodology is particularly relevant for microblogging platforms where it can enable social matching, thus bringing the technology closer to human social behaviour.
Skënderi’s study also demonstrates that traditional text representation methods provide strong baselines that are at their most effective when combined with neural network-based methods. This integration opens new possibilities in the domain of text-based data analysis.
The dissertation offers a vital contribution to the field of text analysis, particularly in the context of Big Social Data. It not only evaluates the applicability of different text presentation methods but introduces a new ensemble learning-based approach, demonstrating the potential to significantly impact the way we manage and interpret textual content generated on digital social platforms.
Erjon Skënderi completed the dissertation in the Doctoral Programme of Computing and Electrical Engineering in the field of Signal Processing and Machine Learning. He is working in the SODA research project at the University of Helsinki where he continues to expand his research.
Public defence on Friday 25 August
The doctoral dissertation of Erjon Skënderi titled Text Representation Methods for Big Social Data in the field of signal processing and machine learning will be publicly examined at the Faculty of Information Technology and Communication Sciences of Tampere University at 12:00 on Friday 25 August 2023 in SA203 (Auditorium S2) of the Sähkötalo building, (address: Korkeakoulunkatu 3, Tampere). The Opponent will be Professor Heri Ramampiaro from the Norwegian University of Science and Technology while Professor Kostas Stefanidis from the Faculty of Information Technology and Communication Sciences and Jukka Huhtamäki from the Faculty of Management and Business will act as the custodes.
The doctoral dissertation is available online.
The public defence can be followed via a remote connection.
@erjon_skenderi
Illustration: Jonne Renvall/Tampere University