Data for: "Question Answering over Knowledge Graphs for Explainable Satellite Scheduling"

Dataset

Description

This dataset accompanies the research paper “Question Answering over Knowledge Graphs for Explainable Satellite Scheduling”. The paper explores the application of Knowledge Graphs (KG) in satellite scheduling and leverages Natural Language Processing (NLP) techniques to generate queries across four categories: Swap Actions, Single Substitution, Verification, and Forward-Backward. Each category focuses on different aspects of satellite scheduling and is supported by custom code to extract information from the KG, provide explanations by combining queries and results, and address the constraints and definitions of the satellite scheduling problem.
The dataset comprises 17 CSV files organised into four categorised folders, with an additional file containing the scheduled data for the Knowledge Graph data structure.

Categorised Folders:

1.Swap Actions
-Swap_actions: Results without human examples provided for generating explanations.
-Human_swap_actions: Results with human examples provided.
-Swap_actions_table_results: Tabular results generated using the custom code by the LLM.
-Code_rerun_Swap_Actions: Number of iterations for each run (repeated three times).

2.Single Substitution
-Single_Substitution: Results without human examples provided for generating explanations.
-Human_Single_Substitution: Results with human examples provided.
-Single_Substitution_table_results: Tabular results generated using the custom code by the LLM.
-Code_rerun_Single_Substitution: Number of iterations for each run (repeated three times).

3.Verification
-Verification: Results without human examples provided for generating explanations.
-Human_Verification: Results with human examples provided.
-Verification_table_results: Tabular results generated using the custom code by the LLM.
-Code_rerun_Verification: Number of iterations for each run (repeated three times).

4.Forward-Backward
-Forward_Backward: Results without human examples provided for generating explanations.
-Human_Forward_Backward: Results with human examples provided.
-Forward_Backward_table_results: Tabular results generated using the custom code by the LLM.
-Code_rerun_Forward_Backward: Number of iterations for each run (repeated three times).

Last, the satellite schedule (sample covering a 24-hour period)

Satellite Schedule Data
Day3: Contains the satellite scheduling data utilised for building the Knowledge Graph.
Date made available20 Jan 2025
PublisherUniversity of Strathclyde
Date of data productionJan 2024 - Dec 2024

Cite this