Stratigraphic Knowledge Graph (StraKG)

The background of this work is the increasing amount of geoscience literature data shared online, including those from governmental agencies, research institutions, and the crowdsourcing encyclopedia. The knowledge graph is an effective way to explore and analyze open-text data. In this work, we designed and constructed a knowledge graph for the field of stratigraphy, called StraKG, to help process records in the Baidu Encyclopedia, a big open-text data resource in Chinese. The files shared in this repository are code for building the layered structure of the StraKG and extracting instance records and relationships from the open text.

StraKG has a two-layer structure, representing ontologies and instances, respectively. At the top is the schema layer for classes and properties of the ontologies. Community-level geological dictionaries were used as a foundation to build ontologies. At the bottom is the instance layer. Text mining techniques were used to analyze open text from Baidu Encyclopedia, extracting instances of strata, rocks, and locations, as well as the relationships between those entities. In our work, we also established mapping between the schema layer and instance layer and implemented a list of experiments to test the utility of the resulting StraKG.