Knowledge modelling and ontologiesTo allow an AI-system to "understand" it surroundings, and the task it should perform, either requires that the system learns how to behave or that we provide this knowledge explicitly to the system. To give the system knowledge about its surroundings through a knowledge model, such as an ontology with associated data in the form of a Knowledge Graph, as well as generic rules of how to act and derive new knowledge from that model, is one way of addressing that challenge. However, this requires that a knowledge model is available, and can be kept up to date. Eva's research targets methods and techniques to create, maintain and evolve knowledge models, in the form of ontologies and Knowledge Graphs.
To create knowledge models is not an easy task, even for an experienced knowledge engineer, and even more challenging for system developers, domain experts, and other participants in the development of an AI system. This is why we need methods and tools to support this process, and which increase the quality of the resulting models. In her PhD dissertation Eva proposed and elaborated the novel notion of Ontology Design Patterns (ODP), as a way of encoding best practice design solutions. During her postdoc she also participated in the development of an agile ontology engineering methodology, eXtreme Design (XD), which uses ODPs to support fast ontology development, as well as iterative ontology evolution, while maintaining a high quality of the model. The method and the patterns have since then been applied and further developed in a number of domains and for several kinds of use cases, such as AI-systems in the security and e-health domains.
Knowledge representation and machine learningClassically, knowledge models and Knowledge Graphs are developed manually, which is a tedious and error prone process, even if tools and design patterns may be a step in the right direction. However, there is already a lot of knowledge available in the form of natural language, e.g. as written text. A part of Eva's research targets to use this knowledge source to automate the modelling process, and thereby generate ontologies and Knowledge Graphs from natural language texts. By using modern machine learning methods and language models, the knowledge in text could be matched to design patterns and thereby modelled appropriately. Resulting Knowledge Graphs will be better tailored for their intended task, and the knowledge model can also be used to reason on the correctness and reasonability of the information extracted from text.
Streaming data and stream reasoningAnother challenge is to manage information that is not static, but becomes available to the system a bit at a time, in a data stream. Examples of this exist in various monitoring scenarios, such as security surveillance, environmental monitoring, as well as e-health and monitoring of patients. Parts of Eva's research targets how knowledge models can be used for analysing and reasoning over streaming data, within an AI system.
Security and e-healthApplications of the research can be found in many different domains, and earlier Eva have worked within various areas, such as the management and publishing of linked geographic data, statistical data, energy efficiency data, and other data from authorities on the web. Recently applications can mostly be found within the security and e-health domains. Eva is a part of the national Security Link research network, and have participated in several EU-funded projects in the security area, such as VALCRI and SPIRIT. Within e-health Eva was previously part of RISE and their activities in e-health, and the now completed project e-care@home.