Autotagging

When you left-click on an existing span annotation, one of your options is "Autotag matches":

If you select this option, all the eligible spans in the text which are not yet marked and match the existing span will be assigned the same annotation label as the already-annotated span.

Default autotagging properties

Autotagging has the following default properties:

• By default, it is case-sensitive; so "George" and "george" will not match. You can control whether autotagging is case sensitive via the "View -> Autotag is case-sensitive" switch.
  
• It normalizes span-internal whitespace; so any sequence of whitespace in your already-annotated span will match any sequence of whitespace in the corresponding position in the target spans.
• It will not annotate in untaggable regions, nor will it annotate any span which is already annotated.
  
• It is sensitive to specific boundaries. If your document is tokenized, the autotagging will only match spans whose boundaries match token boundaries. If your document is not tokenized, the span must be delimited by default by either whitespace or a zone or document edge. Your task maintainer can augment this behavior by specifying the tokenless_autotag_delimiters property, per language, in the task configuration file.

Controlling and enhancing autotagging

There are numerous ways to control and enhance this autotagging capability.

Controlling case-sensitivity

By default, autotagging is case-sensitive, but you can make it case-insensitive by deselecting "Autotag is case-sensitive" in the View menu:

You can also change this setting via "View -> All settings..." in the "Annotation operations" tab, by toggling "Take case into account...":

Controlling untokenized autotagging boundaries

If your document is not tokenized, the characters that determine the edges of autotag candidates via the tokenless_autotag_delimiters property, per language, in the task configuration file. Here, we show the relevant block from the definition of the "Named Entity" sample task:

     <languages>
     	<language code='en' name='English' tokenless_autotag_delimiters='.,/?!;:'/>
     </languages>

So, for instance, if you're using this task, if you annotate "Pakistan" in an untokenized document, and you want to autotag all other "Pakistan" instances as the same label, any such "Pakistan" instance will have to be flanked, on each end, by either whitespace, the beginning or end of the document, or one of the delimiting characters shown above. So the substring "Pakistan" in " Pakistani" will not be a candidate (because "i" is not a whitespace character or one of the delimiters), but the "Pakistan" substring in " Pakistan," will be (because the comma is one of the delimiters). Note that the substring "Pakistan" in "the India-Pakistan border" is also not a candidate, because the dash is not one of the delimiter characters; you could change this by adding the dash to the set of delimiter characters in your task.

Enhanced autotagging operations

You'll notice, in the "Annotation operations" tab above, there's one other setting, to display additional options for autotagging. If you enable this option, your popup menu will contain, instead of "Autotag matches", an "Autotag..." entry with a submenu containing the following operations:

• "Add matches": this is equivalent to the original "Autotag matches" operation.
• "Delete matches": this operation removes all annotations (including the one selected) with the same effective label which marks the same string as the annotation you've clicked on. This operation respects the case sensitivity setting for autotagging.
• "Replace matches with <label>": this operation changes the labels of all annotations (including the one selected) with the same effective label which marks the same string as the annotation you've clicked on. This operation respects the case sensitivity for autotagging.