Gazetteer#

Gazetteer is a user-supplied dictionary mapping labels to terms. Attach one to a Tagger to make the tagger recognize domain-specific vocabulary — product names, code identifiers, internal jargon — that the system model wouldn't normally catch.

#Construction

const gazetteer = new NaturalLanguage.Gazetteer({
  product: ["Scripting", "Pro"],
  company: ["Acme", "Acme Corp"]
}, "en")

• dictionary: { label: [term, term, ...] }. Each label can have many terms; the gazetteer maps any matching term back to that label.
• language?: Language — optional hint when terms are language-specific.

The constructor throws if the dictionary is malformed (e.g. empty terms or duplicate terms under different labels).

#Properties

#Methods

#label(term: string): string | null

Returns the label term maps to, or null for a miss.

gazetteer.label("Scripting")   // → "product"
gazetteer.label("Acme")        // → "company"
gazetteer.label("Banana")      // → null

#Using a gazetteer with Tagger

const gz = new NaturalLanguage.Gazetteer({
  product: ["Scripting"],
  company: ["Acme"]
}, "en")

const tagger = new NaturalLanguage.Tagger(["nameType"])
tagger.setGazetteers([gz], "nameType")
tagger.setText("Acme ships the Scripting app.")

const tags = tagger.tags(null, "word", "nameType", {
  omitWhitespace: true,
  omitPunctuation: true
})
// → tags include { tag: "company", ... } and { tag: "product", ... },
//   instead of just OrganizationName / nil.

Gazetteer hits override the system model for that scheme. If the tagger is constructed with "nameType", then setGazetteers([gz], "nameType") will replace standard OrganizationName / PersonalName labels with your custom labels for the words that match.

#Notes

• This is the lightweight, in-script way to extend tagging. The heavier (and out of scope here) alternative is NLModel, which trains a full classifier from labeled data.
• Multiple gazetteers can be attached to the same scheme via setGazetteers([gz1, gz2, ...], scheme).