xfst-kom.txt test works, but ukom doesn't (

[albbas]

This issue was created automatically with bugzilla2github

Bugzilla Bug 1229

Date: 2011-12-27T21:18:26+01:00 From: Jack Rueter <> To: Trond Trosterud <> CC: rueter.jack, sjur.n.moshagen

Last updated: 2012-02-07T14:51:24+01:00

[albbas]

Comment 5494

Date: 2011-12-27 21:18:26 +0100 From: Jack Rueter <>

In the xfst-kom.txt file there is a testing procedure: ! To test: ! xfst -e "source xfst-kom.txt" ! invert net ! save s ! cat xfst-kom.txt | grep €|cut -d" " -f3|cut -d":" -f1|lookup s

At present this test shows no problems. There are 41 test items.

When make lexfiles is completed, and make has also been done, ukom does not produce the required results for the following rule:

define iPalatalisation [ і -> и || [ Vow | CnsBeforeCyri | ь | д з ] %> _ ] ;

!€ ворссь>і:ворсси !€ ло>і:лои !€ шу>іс:шуис

[albbas]

Comment 5500

Date: 2011-12-28 14:47:53 +0100 From: Jack Rueter <>

This is not an xfst problem, since nouns will work but verb will not. л -> в works in ныл:ныв noun but not кыл:кыв verb

Verbs differ from noun in their lexc in that verbs include serial-verb flags.

[albbas]

Comment 5501

Date: 2011-12-28 15:50:20 +0100 From: Jack Rueter <>

In gtsvn/kt/kom/src I run make lexfiles make

In the second run this is the first of a number of Label rejections:

*** Warning: Label '@U.CONJ-CONNEG.YES@:@U.CONJ-INF.NO@' is illegal: flag diacritics on both sides of the symbol pair.

All of the problems seem to hinge around the fact that there are no interim morphemes. The upper and lower-language pair notation claims to reject flags when they appear on both sides of ":".

[albbas]

Comment 5502

Date: 2011-12-28 16:48:19 +0100 From: Trond Trosterud <>

So, the problem seems to be linked to flag diacritics. Thus, back to reading The Book.

[albbas]

Comment 5747

Date: 2012-02-07 14:51:24 +0100 From: Trond Trosterud <>

After discussion with Lauri, Jaska has solved the problem.