[Run Apertium] See why new Apertium tools give a different chunked result

[rmlockwood]

In the k-z project in the 'ger and -iyam' text for the line GEN 1: 2, I get different results when transferring the word rûyê.
With the old Apertium rules I get:
ch_n NP OBLIQUENESS [rî1.1 n ez 2 m]

With the new Apertium tools I get:
ch_n NP OBLIQUENESS [/î1.1 n ez 2 m]

[mr-martian]

This looks like apertium/apertium#199

This can probably be fixed by changing

${FLEXTOOLS_PATH}\apertium-transfer -t tr.t1x transfer_rules.t1x.bin bilingual.bin ${SOURCE_PATH}

to

${FLEXTOOLS_PATH}\lt-proc -b bilingual.bin ${SOURCE_PATH} | ${FLEXTOOLS_PATH}\apertium-transfer -t tr.t1x transfer_rules.t1x.bin

in the makefile. Invoking apertium-transfer with a bilingual dictionary is deprecated.

[rmlockwood]

This worked for me except for one of my regression tests where the affix looks like <1/3SG>. Maybe apertium-transfer could handle that in the input file and let it go through, but lt-proc doesn't like it and makes a mess of things?

[mr-martian]

blah.dix

<dictionary>
  <sdefs>
    <sdef n="a/b"/>
    <sdef n="x/z"/>
    <sdef n="n"/>
  </sdefs>
  <section id="main" type="standard">
    <e><p><l>a<s n="n"/><s n="a/b"/></l><r>b<s n="n"/><s n="x/z"/></r></p></e>
    <e><p><l>c<s n="n"/></l><r>d<s n="n"/></r></p></e>
  </section>
</dictionary>

$ lt-comp lr blah.dix blah.bin
main@standard 5 5
$ echo '^a<n><z>$ ^a<n><a/b>$ ^c<n><a/b>$' | lt-proc -b blah.bin 
^a<n><z>/@a<n><z>$ ^a<n><a/b>/b<n><x/z>$ ^c<n><a/b>/d<n><a/b>$

Tags with slashes are behaving as I would expect here. What sort of failure are you getting?

[rmlockwood]

Here's some sample data where I get the failure.
Expected result: ^älska1.1$
Actual resuilt: ^3SG>$
issue_612.zip

Maybe the issue isn't in lt-proc, but in apertium-transfer.

[mr-martian]

Ah, yep. When apertium-transfer splits the incoming word into source and target, it checks for / without seeing whether that / is inside a tag or not.

[mr-martian]

Fixed in apertium/apertium@b5916c5 (Apertium 3.9.5)

[rmlockwood]

Will a Windows build with the fix be possible?

…

________________________________ From: Daniel Swanson ***@***.***> Sent: Thursday, May 9, 2024 9:29 PM To: rmlockwood/FLExTrans ***@***.***> Cc: R Lockwood ***@***.***>; Author ***@***.***> Subject: Re: [rmlockwood/FLExTrans] [Run Apertium] See why new Apertium tools give a different chunked result (Issue #612) Fixed in ***@***.***<apertium/apertium@b5916c5> (Apertium 3.9.5) — Reply to this email directly, view it on GitHub<#612 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABWCQCND3OELE53DTJBO5TDZBOB6JAVCNFSM6AAAAABHFCS272VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMBSG43TOMRRHA>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[mr-martian]

An updated windows build process is being worked on, but I have no idea what the timeline for it is like.

[rmlockwood]

this will be fixed with the #708 PR