Hybrid Knowledge Engineering to Build a Global Assassination Dataset

Abstract

Advances in Knowledge Engineering tools such as Named Entity Recognition, Knowledge Graphs, and Machine Learning allow researchers to generate new and more robust linked datasets from which researchers can make new discoveries. It is important for the AI research community to continue leveraging these tools for knowledge discovery while also acknowledging that each tool comes with limitations and an effective scope of use. This paper seeks to highlight the limitations of each of these tools while also uniting each of their strengths to propose a novel methodology leveraging existing databases, knowledge graphs, NER, and ML to build a knowledge repository within the constraints of a small labeled dataset and unstructured and incomplete existing databases. This methodology is implemented to build an enriched assassination dataset that will be made publicly available and assist in future political science research.