Several USC/ISI scientists have used various data-driven methods to automatically harvest large numbers of (typically named) instances from large corpora such as the web. Some of these results have attempted to disambiguate between possible appropriate superconcepts for a harvested instance (consider the difference between a director of a company and a director of a film); others have not.
| Sub-ontology | Namespace | Size | Description |
|---|---|---|---|
| MFI | MFI | 470,000 instance concepts
15,000,000 links + attributes |
Michael Fleischman |
| PPI | PPI | 26,000 instance concepts
315,000 links + attributes |
Patrick Pantel |
| PPI-SNS | 26,000 senses | ||
| PPI-EN | 26,000 English lexical items | ||
| DRI | DRI | 738,000 instance concepts
8,600,000 links + attributes |
Deepak Ravichandran |
| DRI-SNS | 777,000 senses | ||
| DRI-EN | 777,000 English lexical items |