In this position paper, we present our approach and early results in engineering an open research knowledge graph for a scholarly domain. All code, data, and results are publicly available for academic purposes at: https://github.com/tyaroshko/orkg-acd . The domain of Anti-Corruption has been chosen as the use case as it is a vibrantly developing field of scholarly research at the intersection of several fields and research communities, such as Legal, Information Science, Data Analytics, Governance, etc. Furthermore, having a methodologically sound and knowledge-based approach for anti-corruption is in demand in Ukraine on its way toward becoming a member of the European Union. Our approach for building the knowledge graph is based on the use of terminology saturation analysis that ensures the representatives of the used literature sample for knowledge extraction. For building knowledge representations from the recognized terminology, a semi-automated approach with a human in the loop is exploited, including the use of LLMs and zero-shot prompt engineering. The results are further validated using human experts.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Engineering Scientific Knowledge Graphs from Publications: The Anti-Corruption Use Case

  • Taras Yaroshko,
  • Victoria Kosa,
  • Oleksii Ignatenko,
  • Oleksii Makarenkov,
  • Vadim Ermolayev

摘要

In this position paper, we present our approach and early results in engineering an open research knowledge graph for a scholarly domain. All code, data, and results are publicly available for academic purposes at: https://github.com/tyaroshko/orkg-acd . The domain of Anti-Corruption has been chosen as the use case as it is a vibrantly developing field of scholarly research at the intersection of several fields and research communities, such as Legal, Information Science, Data Analytics, Governance, etc. Furthermore, having a methodologically sound and knowledge-based approach for anti-corruption is in demand in Ukraine on its way toward becoming a member of the European Union. Our approach for building the knowledge graph is based on the use of terminology saturation analysis that ensures the representatives of the used literature sample for knowledge extraction. For building knowledge representations from the recognized terminology, a semi-automated approach with a human in the loop is exploited, including the use of LLMs and zero-shot prompt engineering. The results are further validated using human experts.