[Fsa-guatemala] Language and translation of rights

[Daniel Kahn Gillmor]

On Wed 2008-09-17 10:23:42 -0400, Jamie McClelland wrote:

> I'm making one suggested change to our vocabulary. I'm worried that our
> discussion, which is already complicated, is made more difficult to
> follow by using the word "Localization" instead of the more vernacular
> "Translation." I just re-wrote my response using translation in place of
> localization and I found it much easier to follow.

The reason i had explicitly avoided the term "Translation" is that it
implies there is an "Original".  In the context of a Right with two
existing Localizations, I don't believe that the notion of what is an
"Original" is a useful one, and i wanted to avoid it.  Also, when
we're talking about a Right with one Localization missing, there *is*
a sense in which one piece is the Original and the other is the
Translation.  If we're calling them both "Translations", then we can't
distinguish in this case.  It might also be useful to offer
machine-generated translations at some stages in the process (thoughts
on this later).  In this case, i'd use the term "Translation" to refer
to these automated versions, since they are categorically different
than the human-generated (and -discussed) ones.

So i'm continuing to use "Localization" for a language-specific
instantiation of a Right for now, but i'll switch to "Translation" if
other people in the group prefer it.  Other opinions?

> I think we *are* in agreement that participants in the workshop should
> be able to do translations (and this discussion is on how that is to be
> done).

Yes, that's correct.

> I don't think we're in agreement as to whether we should have
> outside translators. Maybe once we've made headway on how
> participants do translations we can go on to discuss whether or not
> to have outside translators.

i'm happy to defer addressing this question, though i would phrase it
as "we have not reached agreement about whether there should be
translators who are not full participants in the process."

> I don't think we should engineer the groups to be multi-language. I
> think we should try to minimize the engineering of the groups. I
> agree - we will most likely have one or probably many groups in
> which there are no fluent spanish speakers or no fluent english
> speakers.

This is fine by me.  does anyone think we should attempt to do group
engineering other than the current method of counting off to "shuffle
the deck" so that people who come in together don't get seated
together?

> I'm working on the same assumption. Unless someone has a vision of
> how a cross-room group could function, I think we should stay with
> the assumption that small group membership will be restricted to a
> single room.

OK by me.

> Ah. I think I see the confusion. I think in my last email I may have
> implied that a group should never endorse a translation in a
> language that they don't understand. Which may have been intepreted
> to mean that groups should endorse the *translation* of the right
> rather than the entire right.
>
> What I meant is: I don't think groups should endorse the translation. I
> think groups should endorse the right, which implies an endorsement of
> all the translations. I don't think anyone should be able to endorse a
> right until there is a completed translation in every language.

I agree that Groups should be Endorsing the Right as a whole.  But
this still presents the same confusion.  If my group contains no
English speakers, and we're endorsing a Right (meaning all
Localizations of the Right), we are by implication endorsing a phrase
in English that we don't understand.  The phrase might be "all power
to the people", or it might be "feed the people to the power plants"
-- we don't know, and we've just effectively endorsed it.

What does this mean for the validity of the outcome of the Workshop?

> The flag should apply to each translation. A right can only be
> edited or endorsed if all translations are completed.

Can a Localization marked "complete" be edited while there are other
"incomplete" Localizations in the same Right?

> This mechanism ensures that the democratic process must wait for
> translations to be made. 

Except that it doesn't, because i can still put in "blah blah blah"
(or a machine-generated translation, or whatever) for the language i
don't understand, and mark it "complete".  It's not necessarily in my
interest to do so, because when a proper translation comes along, the
existing endorsements will get cleared.  This incentive to get a good
translation exists whether or not there is a "complete" checkbox, if
you care about endorsements from the groups who speak the unknown
language.

> The goal is to prevent a situation where five un-translated spanish
> language rights are added and five un-translated english rights are
> added. Then, the english language groups edit and endorse the english
> rights and the spanish language groups edit and endorse the spanish
> rights and there is no cross-language collaboration.

In this situation, a bilingual Group has a lot of power; they're able
to take an existing Right, and use it to gather endorsements from
*both* sets of Groups.  Since (presumably) people want to gather
Endorsements for the Rights they care about, there is already
incentive to make sure the right is readable in both languages.

> I think a simple checkbox is fairly straight forward.

One checkbox per Right, or one per Localization?  With what functional
impact on the process?  Does it make the entire Right unendorseable,
or just the Localization?  Can we explain that impact succinctly
before the session begins, or will people be confused during the
process about what's happening with those checkboxes?  I'm not
convinced yet that it's as simple as a simple checkbox.

> I also think it will encourage groups to attempt translations even if
> they are not confident in their abilities. On the board, I think all
> translations should be displayed regardless of whether or not they are
> complete.

Don't you think that getting Endorsements from Groups with other
linguistic alignments is already incentive enough to solicit improved
translations? 

> Therefore, an english language group with one so-so spanish speaker
> could propose a right in english. The so-so spanish speaker could
> attempt a translation, but leave the completed box unchecked because
> they know it's not quite right. 

This so-so spanish speaker should attempt the Translation (in this
case it *is* a Translation, not a Localization) -- but what about the
scenario where this is an edit of an existing Right?  How does the
"incomplete" flag get used in this scenario?

> In this scenario, the english and incomplete spanish translation
> becomes visible to everyone so it can be discussed in both
> languages, even though the translation is not yet complete (so it
> can't be officially endorsed or edited yet).
>
> It also allows a native spanish speaker with so-so english skills a
> better chance at fixing the spanish translation because they don't
> have to start from scratch.

But if the so-so spanish speaker put in a clumsy attempt and there was
no complete/incomplete checkbox, people could still fine-tune the
proposal (in both languages), the same way that wordsmithing takes
place in the L=1 workshops, right?  Again, i don't see the gain.

> Yes! I think this is right. In fact I think that should be the
> default behavior without any need to indicate it. I think if you
> submit a right or edit a right, your endorsement remains after each
> translation is made. 

How could we do this without gathering some information about what the
fluency level is of each group?

What if bi-lingual Group G fine-tunes the language X Localization of
Right A, completing it, and thereby endorsing Right A?  If the next
revision of Right A modifies only the Language Y Localization, should
G's Endorsement of A stand?  What if the next revision only modifies
the Language X Localization?  Why should it work that way?

> In other words, we do have an implicit trust that translations are
> as accurate as they can be. If you are a bi-lingual group and you
> decide that the spanish translation of your english right is bad,
> then you should be able to edit the right in spanish and indicate
> that the english translation is complete (without changing the
> english translatin). This action will result in losing endorsers,
> but since accuracy in translation is subjective, I think that's
> acceptable.

I'm not sure i understand the proposal here.  First, i believe it was
stated that an Incomplete Right (that is, a Right with one or more
Incomplete Localizations) is unEndorsable.  In this paragraph, it
seems that there is already (at least one) Endorsement on a Right with
an Incomplete Localization.  This seems like a contradiction, no?
What am i missing?

-----------------

Whew!  This is confusing business, particularly because we seem to
need to think about 3 or 4 different stages in the life cycle of each
Right, Localization, and Group.  We'll be implicitly asking the
participants in the Workshop to mentally model the same processes, but
we're giving them much less time to think about it than we have.  This
is one of the reasons i'm afraid of the complexity involved in this
Incomplete flag.

Wow, these are long discussions.  I'll try to follow up with shorter
bits!

           --dkg
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 826 bytes
Desc: not available
Url : http://lists.mayfirst.org/pipermail/fsa-guatemala/attachments/20080917/6aae43d4/attachment.pgp