Representing Scientific Literature Evolution via Temporal Knowledge Graphs

Abstract

Scientific publications register the current knowledge in a specific domain. As new researches are conducted, knowledge evolves, getting documented in dissertations, theses and articles. In this article, we introduce new methods that exploit Temporal Knowledge Graphs (TKGs) to model temporal knowledge evolution in corpora of unstructured texts. In our approach, complex network measurements are applied over TKGs to determine the relevance of concepts dealt with in the corpora under analysis. We demonstrate the effectiveness of our method by conducting experimental analyses on TKGs constructed from a corpus of scientific papers extracted from different editions of the International Semantic Web Conference (ISWC). The results demonstrate the effectiveness of the method in representing and tracking the knowledge evolution over time.