Hi All, The NameFinder implementations create spans based on the entity's token-based start/end indexes.
But, Span.getCoveredText() gets the covered text based on the character start/end instead of the token start/end. An example: sentence = "Neil Abercrombie Anibal Acevedo-Vila Gary Ackerman" span = [0..2) person And getting the covered text of the span: span.getCoveredText(sentence) returns "Ne" and not "Neil Abercrombie" Span.getCoveredText() is used a lot so it seems like changing its behavior is not the fix. Is the problem that the NameFinder implementations are using token positions instead of character positions when creating Spans? Meaning that NameFinderME/DL should be updated to use character start/stop positions instead of token start/stop positions? Thanks, Jeff
