"A credit rationale forms an integral part of the credit granting and credit decision making process. Historically such credit rationales were written as a natural language document. This meant that automated handling of the meaning, i.e.: semantics, of such documents by machines was too complex to be done with sufficient accuracy. As a consequence, the creation and verification of a credit rationale is a time consuming and manual process. The solution that we propose is to create a Linked Data based credit rationale that we call the Advanced Credit Rationale (AdCR). Linked Data can ensure that the semantics in the credit rationale are both human and machine understandable. This enables the automated checking of the credit rationale for potential regulatory issues, as well as semantic based querying over a whole portfolio of such credit rationales. An issue that often hinders the use of a Linked Data based approach is that the creation of the knowledge graph can be challenging for domain experts. In this work we propose an approach that facilitates the creation of a credit rationale that is based on a Linked Data knowledge graph. We evaluate this approach with a prototype. We also demonstrate that with this approach the creation of a Linked Data based credit rationale should not take more time than a natural language based credit rationale. In addition we show that the structures that a Linked Data based credit rationale provides are generally seen as helpful."
Newres Al Haider, Keng Ng, Ali Hashmi, Lauma Veidemane, Diederik Schut
1st Workshop of Knowledge Graphs for Semantics-driven Systems Engineering (KG4SDSE) A Workshop at CAiSE2023 Appears in Advanced Information Systems Engineering Workshops. CAiSE 2023. Lecture Notes in Business Information Processing, vol 482