Package opennlp.tools.namefind

Interface OffsetMappingNameFinder

All Superinterfaces:
TokenNameFinder
public interface OffsetMappingNameFinder extends TokenNameFinder
A TokenNameFinder that can additionally report detected spans in the character coordinates of the original input, mapping back through any text normalization applied before detection.

An implementation that normalizes input before detection (for example an ONNX model that folds Unicode whitespace or dashes) returns spans from TokenNameFinder.find(String[]) in the coordinates of the normalized text, which no longer line up with the caller's input when a fold changes the length. findInOriginal(String[]) maps those spans back to original-input coordinates. This is a separate capability interface rather than a method on TokenNameFinder because the classic contract reports token-index spans, for which an original-character mapping is not meaningful; an interface-typed caller tests for the capability (finder instanceof OffsetMappingNameFinder) instead of depending on a concrete implementation.

  • Method Summary

    Modifier and Type
    Method
    Description
    Span[]
    findInOriginal(String[] tokens)
    Finds names and returns their spans in the character coordinates of the original input, regardless of any normalization applied before detection.

    Methods inherited from interface opennlp.tools.namefind.TokenNameFinder

    clearAdaptiveData, find

  • Method Details

    • findInOriginal

      Span[] findInOriginal(String[] tokens)
      Finds names and returns their spans in the character coordinates of the original input, regardless of any normalization applied before detection.
      Parameters:
      tokens - The tokens to search.
      Returns:
      The detected spans, in original-input character coordinates.
      Throws:
      IllegalArgumentException - Thrown if tokens is null or contains a null token.