Software and information system integration with Knowledge Graph

Abstract

This thesis aims to answer the need for integrating existing information systems, and in more detail, the enterprise application(s) to knowledge graph-based systems, like artificial intelligence related solutions. Information systems can have much domain logic and those execute commands that can be interesting for artificial intelligence-based solutions in many ways. This thesis investigates how to create an easy-to-reproduce pattern to extract data from enterprise application and transforming it into knowledge, by storing it to knowledge graph. Knowledge graph creation with integrity and availability of the data is also investigated. In this thesis the design science approach was used in knowledge graph creation to domain-specific context, and by developing, experimenting, and evaluating the design. Resulting research artifact was used successfully in conjunction with large language models to test the accuracy for answering to 7 questions that were created for the tests. This experiment created a design that makes it possible to create accurate domain-specific knowledge graphs from enterprise applications. These knowledge graphs are also useful for other enterprise applications, that need integration to domain-specific information and data. This knowledge and data are needed for business processes with enterprise applications that need knowledge from other parts of the enterprise architecture.