Bug 2217

Summary: +Pass tag should be changed for sme?
Product: sme lexicon Reporter: Lene Antonsen <lene.antonsen>
Component: Continuation lexicaAssignee: Thomas Omma <thomas.omma>
Status: ASSIGNED ---    
Severity: normal CC: heli1401, lene.antonsen, linda.wiechetek, sandra.rahka, sjur.n.moshagen, tiinapl78, trond.trosterud
Priority: P3 - Within a week    
Version: unspecified   
Hardware: Macintosh   
OS: Other   

Description Lene Antonsen 2016-09-21 07:34:16 CEST
Tiinna wrote an email in July about the problem with +Pass. I add the issue to bz because we have to solve it.

Resume: +Pass is now optional in FSTs because of the use in sme: giving extra information to the grammarchecker for certain verbs (-hallat-verbs):
časkkáhallat	časkkáhallat+V+IV+Pass+Inf


But for other languages, as Tiina mentions in her email, +Pass gives crucial morphological information about the verb, and cannot be optional. 

Suggestion:
* choose another tag for sme, and make it optional
* make the +Pass not-optional



TIINAS EMAIL
The question has come from rus_oahpa needs from one side and sme grammar checker from the other side. 
Both languages use the +Pass tag, but it means different things. In rus language it indicates inflectional passive forms (as opposed to active forms in paradigm) while in sme it is lexical feature that also indicates passive, but on the lexical basis (in addition to morph. tags +Der/Pass indicating systematic passive formation). 

The question came from the fact that +Pass tag was made optional generally for generating of all gt languages. This fits well if it is a lexical feature, that does not need to be given to generator to generate the forms because it does not influence the form generation (as for sme). But it cannot be optional tag if adding or omitting it will result in different forms (as for rus, where optional behaviour generated both active and passive forms where it was supposed to be only active one - as an answer to an active question in oahpa).

My suggestion is that if this tag means different things and it is desired that it will be functioning on a general level - the same way for all gt languages - then it should be split into two different tags so one of these usages should be renamed. Then one of them can still function (be optional for generators) on a general level and the other will not (as it is undesired behaviour). 

The issue is relevant also for other gt languages, all gt languages that currently use the +Pass tag are: 
nob
rus
sme
smn
sms
fin
est
vro
fkv
fao
mhr
vep

I can say that also at least fin, est, vro are using this tag as indicating standard inflectional passive forms. I hope you know best about the other languages and their usage of this tag.  

Best regards,
Tiina
Comment 1 Trond Trosterud 2016-09-21 08:14:35 CEST
I want to follow up on this: Yes, we do not want to have this classification Pass usage. 

There are e.g. sme-fin and fin-sme MT systems, and we do not want to have the same tag meaning different things at both sides of the fence in those systems (and of course not having the tag be facultative). Also, given that the sme +Pass tag reports verbal behaviour (much like e.g. the distinction between "real" TV and "mostly TV" verbs, e.g.) the tag need a prefix saying that this denotes usage of behavior X. +Use/X may be too loaded (it denotes __our__ use, and not __language behaviour__, so perhpas +Behaviour/, +Type/, +Kind/, or something of this kind could be a solution. The tag should probably be deleted outside grammarchecker use, rather than being done facultative? 

But distinguishing such verbal behaviour in itself seems a good thing, as soon as we finds a less misleading way of doing it.
Comment 2 Linda Wiechetek 2016-10-10 08:49:56 CEST
passive, causative, reciprocal, reflexive verbs are interesting valency-wise as they give us information about the number and type of arguments to some degree. However, in cases of ambiguity, i.e. a reflexive verb that can also be continuative etc. we want to be able to keep track of that because that changes the valency potential.
Are there any passives that are ambiguous with other derivations? Maybe Duommá or Lene could answer that question.
So my point is that if we include passive, we might have to think about tags for the other derivations too (which could be nice).

As a prefix I suggest something refering to Alternation (Alter) or Diathesis (Dia).
Comment 3 Sjur Nørstebø Moshagen 2016-11-17 11:07:37 CET
See also the discussion in bug #2245.