Bug in trimming?
I get this:
$ echo "jouer du violoncelle" | apertium -d . fra-oci
@jouer lo violoncèl
To analyse the issue, I do:
$ echo "jouer du violoncelle" | apertium -d . fra-oci-disam
"<jouer du>"
"jouer# de" vblex inf SELECT:1514
"le" det def m sg
; "jouer" vblex inf SELECT:1514
; "de" pr
; "le" det def m sg
"<violoncelle>"
"violoncelle" n m sg
"<.>"
"." sent
The problem is that I don't have jouer# de in the bidix, although it exists in apertium-fra. Instead in the bidix there is jouer# de la flûte and jouer# de la trompette:
$ grep "r>jouer<" .deps/fra-oci.dix
<e r="LR"><p><l>flaütar<s n="vblex"/></l> <r>jouer<g><b/>de<b/>la<b/>flûte</g><s n="vblex"/></r></p></e>
<e r="LR"><p><l>flaütejar<s n="vblex"/></l> <r>jouer<g><b/>de<b/>la<b/>flûte</g><s n="vblex"/></r></p></e>
<e> <p><l>jogar<s n="vblex"/></l> <r>jouer<s n="vblex"/></r></p><par n="v-v_tv"/></e>
<e r="LR"><p><l>jugar<s n="vblex"/></l> <r>jouer<s n="vblex"/></r></p><par n="v-v_tv"/></e>
<e> <p><l>protagonizar<s n="vblex"/></l> <r>jouer<g><b/>le<b/>rôle<b/>principal</g><s n="vblex"/></r></p></e>
<e a="hector"><p><l>trompetar<s n="vblex"/></l><r>jouer<g><b/>de<b/>la<b/>trompette</g><s n="vblex"/></r></p></e>
<e a="hector"><p><l>tamborinar<s n="vblex"/></l><r>jouer<g><b/>du<b/>tambour</g><s n="vblex"/></r></p></e>
I suspect that the trimming messes up and includes jouer# de, which causes the error.
Since you're going in the fra->oci direction, wouldn't trimming be about whether jouer was present on the left side of the bidix?
You are probably right. When I saw @ in the output I automatically associated to a trimming problem, since @ errors disappeared after trimming was included.
The fact is that I get an unusual error, and a morphological analysis with a word that is not in my bidix. jouer# de should not appear at all in the analysis, as any word in the monodix should not appear if it is not used in the bidix.