The 3rd TranSEHA Seminar

Dr. Qi Zhang（ROIS-DS・Project Researcher）

【Title】
Leveraging Knowledge Graphs for Interdisciplinary Data Integration in Polar Science

【Abstract】
Polar science draws on data from many domains, including biology, oceanography, meteorology, cryospheric science, and geology. These data are published across numerous databases and web resources. Although access to individual data resources has improved, it remains difficult to identify records relevant to a research question across disciplinary and institutional boundaries and to understand the scientific connections among them.

To address this challenge, we are developing the Polar Science Knowledge Graph (PSKG) as a record-level graph index spanning distributed data sources. PSKG has a three-layer architecture. The source layer represents the original databases and web resources; the data layer contains record-level nodes with links back to the original resources; and the knowledge layer represents reusable scientific entities, such as organisms, regions, and observed variables, together with explicit relations among them. We are also developing PENGUIN, an exploration platform built on PSKG. Through keyword search, interactive graph exploration, and AI-assisted exploration based on natural-language questions, PENGUIN helps users discover relevant records, inspect how they are connected, trace exploration paths and data provenance, and access the original data sources.

The current graph integrates records from ten data sources and contains 17,166 nodes and 74,483 relationships. PENGUIN is publicly available at https://penguin.nipr.ac.jp/. Starting with polar science, we plan to expand the framework to additional domains and develop it into a more broadly applicable infrastructure for cross-disciplinary data integration and discovery. Future work will also improve data updating and quality control and extend support for data analysis and visualization.

Figure 1: Three-Layer Architecture of PSKG

Three-layer architecture of PSKG. The source layer represents existing databases and web resources; the data layer represents records from these sources and preserves links to their original locations; and the knowledge layer represents scientific entities, such as organisms, regions, and observed variables, together with their relationships. By connecting the three layers, PSKG enables cross-disciplinary and cross-platform data exploration while preserving access to the original sources.

Figure 2: PENGUIN Workspace

PENGUIN workspace. Users can perform keyword searches and AI-assisted exploration based on natural-language questions, expand relevant records in the graph, examine scientific relationships, exploration paths, and data provenance, and access the original data sources.

PENGUIN: https://penguin.nipr.ac.jp/