A named entity is a word or n-gram that identifies a specific real-world person, place, thing, date, location, website, company, or other information-bearing entity. Recognition turns text into structured data that downstream systems can use. section 11.3 section 11.4
Common entity categories include:
- People, including pronouns or words whose meaning must be resolved to a particular user. section 11.3
- Places, dates, and events such as Monday, which can be located on a calendar. section 11.3
- Specific things such as websites or companies, whose spelling may require normalization. section 11.3
- Numerical information such as GPS locations, dates, prices, and numbers. section 11.4
Turning a reminder into an action
In “Remind me to read AI Index on Monday,” a useful system identifies “me” as a person, resolves the relevant user, recognizes aiindex.org as a website, identifies Monday as an event, extracts the relation between “me” and “remind,” and selects the next upcoming Monday. section 11.3
Named entity recognition can use pattern matching or neural methods. Regular expressions may be more precise for predictable data such as dates, prices, numbers, and GPS locations; recognized entities can then populate a knowledge graph or trigger an application action. section 11.4
