Back to search

Ελληνικό Μεσογειακό Πανεπιστήμιο

Workload based summaries for knowledge graphs

Abstract

dc:description

This dissertation delves into the advancements in semantic summarization and user centric exploration of Knowledge Graphs (KGs), driven by the rapid expansion of interconnected data. Semantic summaries have become critical tools for distilling vast datasets into manageable sizes, optimizing query answering, indexing, and visualization. Recent developments in structural semantic summaries have focused on extracting central nodes from the semantic graph, exploiting several graph centrality measures, then linking them and presenting them as summaries. Those summaries can then be used among others to optimize query answering as the size of the graph is drastically reduced. However, as the semantic graphs are heterogeneous, using variations of centrality measures for selecting parts of the graph to be used as a summary, generates summaries with limited benefits for query answering. Leveraging user query workloads has the potential to offer tangible benefits to this direction, as they can offer unique insights on trends and user interests as they evolve over time. To this direction, this dissertation starts exploring workload-based summaries by selecting nodes based on their frequency in query workloads. This drastically improves the usefulness of the result summaries in terms of query coverage. Then it explores how utilizing query logs and Large Language Models (LLMs) can lead to the automatically generation of FAQs, enabling users to rapidly understand the contents of an entire KGjust visiting a set of questions and their answers in textual format. In parallel, we explore shifts in user interests over time using query logs and language models, facilitating users' visualization and understanding of these evolving interests. Then we focus on how to construct personalized summaries that adapt to individual user preferences. We again exploit query logs selecting queries similar to the interests of the user and generate summaries maximizing coverage for user queries, dominating all baselines and competitors. Finally we focus on how workload-based summaries can be used for the generation of compact structures that can be used as a caching mechanism to rapidly provide a first answer to user queries before answering their queries in full. We demonstrate that such summaries are both practical, as they can be trivially constructed and retained in main memory, and also of high benefit as they can significantly optimize the time required for the first results of user queries. By integrating these innovative approaches, this dissertation aims to advance the field of semantic summarization and user-centric KG exploration, fostering more effective and efficient data exploration in increasingly interconnected environments.

Degree

thesis:*
Grantor dc:publisher
Ελληνικό Μεσογειακό Πανεπιστήμιο
Year dc:date
2024

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Vasileiou, Ioannis
  • Βασιλείου, Ιωάννης

Subjects

dc:subject × 14

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/57928

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Vasileiou, Ioannis; Βασιλείου, Ιωάννης. Workload based summaries for knowledge graphs. Ελληνικό Μεσογειακό Πανεπιστήμιο, 2024. http://hdl.handle.net/10442/hedi/57928