Dark Knowledge: Knowledge Graphs of What Cannot Happen or Should not Be Done

Oleg Aleksandrovich Sychev

Keywords: Knowledge graphs, Hybrid reasoning

Abstract: While a significant number of large knowledge graphs is developed during the last years, they are mostly used for information search. The ability to reason conclusions on them is limited, and a lot of works turn to approximate methods like neural networks and graph embeddings to draw conclusions and use the accumulated knowledge. In this work, I argue that in human thinking the main usage of logical reasoning is not making far-fetched strict logical conclusions but weeding out obviously wrong ideas and false information while generating new ideas is mostly intuitive. This can be used in constructing hybrid reasoning systems where neural networks play the role of intuition (generating possible solutions) while logical reasoning is used to verify and filter these solutions. This requires creating knowledge graphs containing negative information: the information of what cannot happen or should not be done and why. The problems of creating negative knowledge graphs are discussed. Hybrid systems using negative knowledge graphs will be a lot more trusted as they will have a system of human-verifiable rules that guarantees avoiding the worst errors and filter possible solutions which can be used in many fields from decision making to natural-language parsing.

Subtrack: Problems to Solve Before You Die

First Author Is Student: No