[en] Enhance the disambiguator to allow the 'hiding' of text

[MikeUnwalla]

Some possible users of LT cannot use it because LT is for plain text only. We have some integrations, but they are very limited (https://dev.languagetool.org/software-that-supports-languagetool-as-a-plug-in-or-add-on).

To increase the potential user base (and to make my ASD-STE100 checker available to a wider audience), I want to ignore XML and HTML code and let LT analyse the text AS IF the code did not exist. I cannot do this with immunization (https://forum.languagetool.org/t/what-is-the-purpose-of-immunization-what-should-it-do-or-not-do/9204).

Disambiguator enhancement: Make something similar to 'Immunizing words from matching' (https://dev.languagetool.org/developing-a-disambiguator#immunizing-words-from-matching). But, let it fully ignore the text, as if the text did not exist. This would let users add rules like this to the disambiguator:

  <rule id="SIMPLE_IMMUNIZE_XML" name="Immunize XML code">
      <pattern>
        <marker>
          <token regexp="yes">&lt;</token>
          <token spacebefore="no" regexp="yes">[a-z]+</token>
          <token spacebefore="no" regexp="yes">&gt;</token>
        </marker>
      </pattern>
     <disambig action="hide"/>
    </rule>

Then, for example, with this sentence:
I am in a <em>trouble</em>.
the grammar rule IN_A_TROUBLE would give a warning for: a <em>trouble

Optional extras:

• Show the disambiguation rule name in Tagger Result.
• For testing, permit untouched and ambiguous examples in the rule.
• Automatically ignore the spelling of 'hidden' text.

[danielnaber]

I understand, but a cleaner solution would be to use a pre-processing that removes the (e.g.) XML or use the API's data parameter to send information about what is text and what is markup.

[MikeUnwalla]

@danielnaber, thanks.

1. If the tags are removed, the workflow is 1-way only. For example, I cannot use File>Open in the GUI, do the analysis, correct errors, and then save the text.
2. I do not understand what you mean by "the API's data parameter". Is there a page on the LT website that explains?

Thanks.

[danielnaber]

@MikeUnwalla I was referring to the HTTP documented at https://languagetool.org/http-api/#!/default/post_check - it can receive information about which parts of the input are text and which are markup (but that information needs to be generated by the caller of the API).

[MikeUnwalla]

@danielnaber, thank you.