Entity disambiguation in natural language text
US9245015B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Mar 8, 2013 |
| Grant date | Jan 26, 2016 |
| Priority date | — |
| Expiry date | Jan 23, 2034 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06F40/30
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A device analyzes first text to identify a pair of terms, within the first text, that are alias terms. The device analyzes the first text by performing two or more of: a latent semantic analysis of the pair of terms, based on the pair of terms being associated with a particular tag; a tag-based analysis that determines that the pair of terms are associated with compatible tags; a transitive analysis that determines that a pair of neighbor terms, associated with the pair of terms, are associated with compatible tags; or a co-location analysis based on a distance between the pair of terms in the first text. The device generates, based on analyzing the first text, a glossary that includes the pair of terms identified as alias terms. The device replaces terms within the first text or a second text that is different from the first text, using the glossary.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.