[ar] generate multiple suggestions

Development
1

Hi,
I want to generate suggestion for this case:
I have word like Aqlà => Intlq (in arabic أقلع to انطلق),
due to the rich morphology, this example needs suggestions like this

    if Aqlà:
    suggests: intlq
    if tqlà: suggests tntlq
    if yqlà: suggests yntlq
    if nqlà: suggests nntlq

I tried to generate suggestions like this

<rule>
 <pattern>
  <marker>
   <token inflected="yes">Aqlà</token>
  </marker>
 </pattern>
 <message>
 Suggestions: 
  <suggestion><match no='1' regexp_match="Aqlà" regexp_replace="intlq"/></suggestion>
  <suggestion><match no='1' regexp_match="qlà" regexp_replace="ntlq"/></suggestion>
 </message>
 <example correction="intlq" type="incorrect"><marker>Aqlà</marker> word </example>
 <example correction="yntlq" type="incorrect"><marker>yqlà</marker> word </example>
</rule>

But it generates unsuitable suggestions.
can you help me?

2

Do you really want to replace one word with one other word, not considering context? Or was this just an example? Are these words (tqlà, yqlà, …) forms of Aqlà?

3

I simplify the example to explain how to problem occurs.

yes, as prefixes t-qlà, y-qlà for A-qlà.

4

I might be missing something, but I think you need a <filter> for these cases, i.e. Java programming is required (Development Overview - LanguageTool Wiki).

5

Thanks,
I reformulate my question:

  • How can I replace a ‘verb1’ by a verb ‘verb2’ with the same postag for the conjugated form.
    thanks
6

You can do that like in this example:

<rule id="ELUDED_TO" name="eluded to (alluded to)">
    <pattern>
       <token inflected="yes">elude</token>
       <token>to</token>
    </pattern>
    <message>Did you mean <suggestion><match no="1" postag="VB.*">allude</match> to</suggestion>?</message>
    <example correction="alluded to">He <marker>eluded to</marker> engagement at the party.</example>
</rule>
7

Hi,
I tried this example, I make one with arabic words,

<rule id="bldmn" name="bld mn (blad mn)">
    <pattern>
       <token inflected="yes">blad</token>
       <token>mn</token>
    </pattern>
    <message>Did you mean <suggestion><match no="1" postag="N.*">bld</match> mn</suggestion>?</message>
    <example correction="bld mn"> <marker>blad mn</marker>.</example>
    <example correction="albld mn"><marker>alblad mn</marker>.</example>
</rule>

by testing this example I had :

Incorrect suggestions: bld mn != (bld) mn on input: He blad mn engagement at the party.)
 Incorrect suggestions: albld mn != (bld) mn on input: He alblad mn engagement at the part

Is there any configuration or module to activate, please note that we have an arabic_synth.dict file.
Thanks.

N.B. The given examples are transliterated from Arabic, the original is:

<rule>
    <pattern>
       <token inflected="yes">بلاد</token>
       <token>من</token>
    </pattern>
    <message>Did you mean <suggestion><match no="1" postag="N.*">بلد</match> من</suggestion>?</message>
    <example correction="بلد من"> <marker>بلاد من</marker>.</example>
    <example correction="البلد من"><marker>البلاد من</marker></example>
</rule>

Incorrect suggestions: بلد من != (بلد) من on input: He بلاد من engagement at the party.)
 Incorrect suggestions: البلد من != (بلد) من on input: He البلاد من engagement at the part
8

Solved.
I found an error on arabic_tags.txt file,
it must contain all possible postags.
Thanks