Browsing by Author "Hyvönen, Eero, Prof., Aalto University, Department of Computer Science, Finland"

Filter results by typing the first few letters
Now showing 1 - 5 of 5
  • Building Ontology and Data Infrastructure for Semantic Web Applications
    (2023) Frosterus, Matias
     School of Science | Doctoral dissertation (article-based)
    Semantic Web and Linked Data are paradigms aiming for more efficient and more sophisticated knowledge management through machine-understandable semantics and functional links between entities and relations present in data. Central to this idea is the concept of ontologies: knowledge structures explicitly describing the concepts of a certain domain in a machine-processable way. This dissertation explores ways of developing ontologies from existing thesauri which have been used to produce high-quality metadata for a very long time and how those ontologies can be used in information retrieval. The main research methods used are design science and action research. Extensive application to practice has been used to demonstrate the viability of the proposed solutions. The key results in the dissertation comprise of methods for creating and maintaining a linked open cloud of ontologies and then utilizing those ontologies in linked data publication and for improving semantic search. The primary focus is on data that has been annotated using legacy thesauri that then have been converted into ontologies and the benefits afforded by that. The semantic search techniques employ document expansion using the relations between concepts provided by the ontologies and improvements to information retrieval results are demonstrated. Primary domains explored are related to cultural heritage but the dissertation includes a case study where the methods are implemented for juridical data, as well. The research pioneered in this dissertation has been widely adopted into use in Finnish libraries, archives, museums, and public administration through the national Finto ontology service maintained by the National Library of Finland.
  • Contributions to Self-Organizing Networks and Network Measurement Data Management
    (2019) Apajalahti, Kasper
     School of Science | Doctoral dissertation (article-based)
    In the future, the mobile network infrastructure needs to facilitate wireless communication to automate industry processes in many vertical domains. The heterogeneity of domains and use cases need to be addressed by various traffic service types that require new technologies. The management architecture of the upcoming 5G networks should cover flexible cross-platform optimization (both technology and administrative domains) and operator business objectives. The new management aspects combined with the increasing complexity of the mobile network infrastructure denote the necessity of the adaptive automation of operability and management. Along with the 4G, the Self-Organizing Networks (SON) paradigm has been designed and utilized to automate some network management use cases. The challenge in the future network management is the interplay of cross-platform management functionalities in complex 5G networks. Some of the objectives in the network management for operators in deciding the right context-specific solutions are: 1) comparing similar cross-platform SON functions and their configurations, 2) providing linkages of metrics across platforms, 3) providing graphical user interfaces to understand the decisions and actions of autonomic SON functions, and 4) automating the process of modelling SON functions and their metadata. This thesis is conducted by designing, implementing, and evaluating frameworks, models, and methods, that address the aforementioned challenges. The research follows the principles of the design science methodology. User interface functionalities with faceted browsing activities are designed in order to provide flexible information exploration for the user. The other user interface design offers an interactive SON function discovery mechanism for a prototype SON service system and the other provides an ontology-based visualization of the functionality of an individual SON function. The thesis presents semantic models for reasoning-based SON function discovery and composition mechanism and for defining metric dependencies. Statistical methods are developed for mining time series-based event patterns as context-specific metadata for SON functions and for matching network metrics across heterogeneous datasets with a correlation-based method. All the contributions are reflected against the related work and discussed from the viewpoint of practical benefits for network management. The contributions are novel in view of adapting methods from other research areas to the SON and network measurement data management.
  • From Text to Knowledge: Methods, Tools, and Applications for Digital Humanities Based on Linked Data
    (2023) Tamper, Minna
     School of Science | Doctoral dissertation (article-based)
    The digitization of Cultural Heritage collections has enabled the use of computational methods such as Natural Language Processing (NLP) on textual collections. These methods have been used widely in Digital Humanities (DH) to study digitized contents with automated processes. The Semantic Web and linked data technologies have been applied to describe document collections and their metadata in library and museum collections. They provide infrastructure for connecting different collections by linking them using shared vocabularies that describe metadata values and fields. Linked data is also used in Finnish museum and library collections. It is commonly used to modeling document metadata, such as author, or title of a piece of work. Also, the content of a document in a collection is usually described using manually assigned keywords. Other information about the content is often scarce and finding documents related to an actor can be laborious. This thesis studies and presents novel models, methods, and tools for transforming and enriching document collections automatically to linked data. Linked data technology helps to link together documents of a collection based on their metadata, e.g., author, or publisher. It can be also used to link documents based on information extracted about the content, such as actors mentioned in text. The aim of this thesis is to study how the NLP methods and linked data can be used to study digitized document collections, such as biographies. Research in this thesis is conducted by designing, implementing, and evaluating proof-of-concept systems, tools, and data for real life use cases. The research follows the principles of the design science and action research. The thesis presents a toolkit that can be used to model, transform, and enrich biographical text document collections to linked data to improve collection's information retrieval and interoperability internally and with other collections. The data model for describing text document collection's content and features, e.g., keywords and mentioned names, creates a foundation for building intelligent services based on the linked data such as network or linguistic analysis. These services can be used to visualize the interlinked data by showing the relations between themes or actors. In addition, the linked-data-based datasets can be used as an input for NLP tools to create data analytical visualizations and applications. This approach can be also used to evaluate the quality and content of text document collections for DH research. The prototypes created for data transformation, enrichment, and information visualization can be also applied to other document collections.
  • Modeling and Using Biographical Linked Data for Prosopographical Data Analysis
    (2024) Leskinen, Petri
     School of Science | Doctoral dissertation (article-based)
    Biographical data is used for identifying people, groups, and organizations and for conveying information about them. Biographical data describes life stories of people with the aim of getting a better understanding of their personality, actions, and interperson relations. The underlying texts can also be used for data analysis and distant reading once the documents are provided in a machine-readable format. Prosopographical analysis delves into the life stories of individuals within a defined group to identify shared characteristics and patterns. This dissertation presents and utilizes a comprehensive framework for managing and analyzing biographical data in Digital Humanities research. It includes data models, methods, and applications that enrich biographical content with links and reasoning to enhance the findability, accessibility, interoperability, and re-usability following the FAIR principles. Furthermore, the framework includes versatile tools for both individual biographical research and prosopographical research on groups of people. Linked Data together with event-based data model schemas are used in the published datasets to achieve the interoperability of heterogeneous data regarding historical people. Events are used as the glue combining information from various sources. The event-based modeling enables depicting historical narratives as data, which can be further enriched with the events of individual people and organizations. The research included in this dissertation follows the principles of the design science and action research. The research has been carried out in multiple research projects concentrating on biographical data: WarSampo (2015–), BiographySampo (2018–2021), Norssi High School Alumni (2017), AcademySampo (2019–2021), LetterSampo (2020–2022), and ParliamentSampo (2021–). The data publications and services, online portals, and published articles with analysis are represented as the results of the work accomplished for this thesis. Besides, this thesis tackles the practices of creating, modeling, and publishing Linked Data, as well as analyzing this biographical and prosopographical data by the means of network and data analysis.
  • Ontology Services for Knowledge Organization Systems
    (2017) Tuominen, Jouni
     School of Science | Doctoral dissertation (article-based)
    Ontologies and other knowledge organization systems, such as controlled vocabularies, can be used to enhance the findability of information. By describing the contents of documents using a shared, harmonized terminology, information systems can provide efficient search and browsing functionalities for the contents. Explicit descriptive metadata aims to solve some of the prevailing issues in full text search in many search engines, including the processing of synonyms and homonyms. The use of ontologies as domain models enables the machine-processability of contents, semantic reasoning, information integration, and other intelligent ways of processing the data. The utilization of knowledge organization systems in content indexing and information retrieval can be facilitated by providing automated tools for their efficient use. This thesis studies and presents novel methods and systems for publishing and using knowledge organization systems as ontology services. The research is conducted by designing and evaluating prototype systems that support the use of ontologies in real-life use cases. The research follows the principles of the design science and action research methodologies. The presented ONKI system provides user interface components and application programming interfaces that can be integrated into external applications to enable ontology-based workflows. The features of the system are based on analyzing the needs of the main user groups of ontologies. The common functionalities identified in ontology-based workflows include concept search, browsing, and selection. The thesis presents the Linked Open Ontology cloud approach for managing and publishing a set of interlinked ontologies in an ontology service. The system enables the users to use multiple ontologies as a single, interoperable, cross-domain representation instead of individual ontologies. For facilitating the simultaneous use of ontologies published in different ontology repositories, the Normalized Ontology Repository approach is presented. As a use case of managing and publishing a semantically rich knowledge organization system as an ontology, the thesis presents the Taxon Meta-Ontology model for biological nomenclatures and classifications. The model supports the representation of changes and differing opinions of taxonomic concepts. The ONKI system and the ontologies developed using the methods presented in this thesis have been provided as a living lab service http://onki.fi, which has been run since 2008. The service provides tools and support for the users of ontologies, including content indexers, information searchers, ontology developers, and application developers.