Detection of missing entities in a graph schema
US10789296B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Jul 10, 2018 |
| Grant date | Sep 29, 2020 |
| Priority date | — |
| Expiry date | Nov 15, 2038 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06F18/23213
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
Detecting missing entities in a graph schema is provided. Unlabeled and unstructured data of a particular domain are divided into a plurality of Voronoi cells using a clustering algorithm that is initiated with cluster centroids proportional to a number of entity types corresponding to the particular domain existing in the graph schema. One additional cluster more than the number of entity types corresponding to the particular domain existing in the graph schema is initialized using a cluster initializing formula of the clustering algorithm. It is determined whether the one additional cluster is populated. In response to determining that the one additional cluster is populated, an entity type is determined to be missing from the number of entity types existing in the graph schema. The missing entity type is added to the graph schema.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.