Abstract
Hierarchical relationships between attributes in knowledge representation can have a number of benefits such as automatic alignment of attributes of differing semantic granularity, simultaneous assertion of semantically equivalent properties defined on different levels of a semantic hierarchy, etc. Many knowledge representation systems lack systematic hierarchical alignment of attributes. Since there are few existing sources of such inventories, there exists a bootstrapping problem for superproperty curation. Given the large number of pre-existing properties that can exist in a knowledge graph, it is difficult for human curators to evaluate all of the possible pairings of properties that may potentially represent a superproperty relationship. This disclosure describes a human-in-the-loop machine learning approach to seed a pre-existing collection of knowledge representation schema with superproperty relations.[1]
[1] All authors have contributed equally.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Rezk, Martin; Robinson, Stuart; and Taylor, Edwin James, "Extracting Superproperty Candidates Using Vector Embeddings", Technical Disclosure Commons, (June 25, 2021)
https://www.tdcommons.org/dpubs_series/4416