Comparing Foundations: Insights into the Construction of Financial Causal Knowledge Graphs with and without Ontology

抄録

Simplicity in information navigation, interpretation, and reasoning has positioned mono-relation knowledge graphs (KGs) as a focus of attention, particularly in targeted scenarios. However, their deficiency in hierarchical interactions between entities, such as is-a and instance-of relations, crucial in most commonsense KGs, is a notable limitation. This paper exemplifies mono-relation KGs through FinCaKG - Financial Causality Knowledge Graph and aims to assess whether integrating ontology into FinCaKG can augment their performance in practical applications. For a systematic comparison, we begin by investigating the different methodologies with and without ontology during FinCaKG generation processes. After obtaining both types of knowledge graphs, we carry out an extensive analysis of graph statistics and causality knowledge representations. Our discovery reveals that while mono-relation KGs adequately serve information navigation needs, integrating ontology expands the scope among queried entities.