tlhIngan-Hol Archive: Mon Oct 26 17:56:10 2009

Back to archive top level

To this year's listing

[Date Prev][Date Next][Thread Prev][Thread Next]

Re: Ditransitive reflexives

Tracy Canfield ([email protected]) 

I appreciate the grammatical advice, although I strongly disagree with you
about machine translation.
While I won't claim that the state of the art in machine translation
remotely approaches what a human translator - in the current context,
perhaps it's better to say "a living translator" - can do, I wouldn't say it
requires us to go from expression to thought.  Rather, I'd say that work in
machine translation concentrates on finding an appropriate level of analysis
for the source language that allows a process of transfer to an input that's
suitable for the target language - whether that's a morphological analysis,
a syntactic analysis, a semantic analysis, or something else - and that in
many cases, the results can be quite good.

In the case of Klingon, I would say there are, at the very least, a very
large body of sentences that can be translated by computer into good English
- large enough to make the project worthwhile.  (Handling articles like "a"
and "the", which have to be inserted, is probably going to be the biggest
stumbling block to a natural-sounding English text; dealing with aspect is
probably going to be the biggest stumbling block in analyzing the Klingon.)
So far I've been getting good results using a phrase-structure grammar with
feature unification for analysis.

I am a PhD student in computational linguistics, and this is a class project
for me - though ideally I'd like to end up with something robust enough to
be able to share with the larger Klingon linguistics community.  I won't
claim that I'm going to produce a 100% solution to the problem of machine
translation - that would be folly, given that no one has done it for _any_
language pair - but I do think that 90+% solutions are both possible and
worthwhile.

If there's some sort of gap in our knowledge of Klingon, I'm perfectly
willing to consider certain sentences as untranslatable.  What I want to
avoid is either accidentally including impermissible sentences, or (insofar
as possible) excluding permissible ones.

2009/10/26 Doq <[email protected]>

> Computer translation is extremely difficult because one language is
> not a code for another language. At essence, each expression begins
> with a thought, and the language guides how any given thought will be
> expressed. Translation requires us to go from expression back to
> thought and then to a new expression with different grammars guiding
> us to a new form of that expression. Unfortunately, to whatever extent
> that you can get a computer to speak English or Klingon, you can't get
> it it speak "thought". People can't even speak "thought". They
> certainly can't teach computers how to speak it.
>
> Taking the "simple" example of "They gave chocolate to each other,"
> with a lot of experience using the language, I don't have a simple
> formula that I know for expressing it. I can express it, but I have to
> think about it first.
>
> While Klingon grammar has less redundancy than most languages,
> redundant use of nouns is more common in it than most languages. I'd
> say something like {ghomvamvaD yuch nob ghomvam.} I'm choosing {-vam}
> instead of {-vetlh} simply because there could be two different groups
> referred to as {ghomvetlh}, but there can typically be only one group
> referred to as {ghomvam}.
>
> I could also identify the "same" group in other ways, like {yuch
> nobwI'pu'vaD yuch lunob}. I'm sure there are other ways, as well.
>
> Meanwhile, I think you and others have fallen into the tempting trap
> of trying to use {-chuq} and {-'egh} simply because the phrase "each
> other" is right there staring at us and that's how we say it in
> English. I know of no way to apply these suffixes to anything but the
> subject. It doesn't work with the Klingon equivalent of a
> prepositional phrase. If the noun you want to point it to has a type 5
> noun suffix on it, you can't get there from here. Reflexive verb
> suffixes are only to be applied to nouns that are acting as subject
> and direct object.
>
> In My Humble Opinion.
>
> Doq
>
> On Oct 26, 2009, at 5:04 PM, Tracy Canfield wrote:
>
> > Dear distinguished Klingonists:
> > I'm putting together a machine translation project in Klingon, and
> > have some
> > grammar questions that TKD doesn't seem to cover.  I'm not trying to
> > translate any particular phrase - I just want to make sure my
> > grammar has
> > the fullest possible coverage.
> >
> > In English, we can combine a ditransitive verb like "give" with
> > either a
> > direct object or an indirect object anaphor.  It doesn't seem
> > obvious to me
> > how to do this for certain cases in Klingon.  (I've Googled the
> > archives,
> > but didn't find anything.)
> >
> > First, the reflexive indirect object - "They give each other
> > chocolate."
> >
> > To make a long-form sentence, we need something that can take the
> > dative
> > ending -vaD.
> >
> > ???-vaD yuch nob
> >
> > If we just put "chaH" here, we have a legal sentence:
> >
> > chaHvaD yuch nob
> >
> > but it's indistinguishable from "They give them chocolate", where
> > "they" and
> > "them" refer to different people*.  There isn't a stand-alone
> > pronoun that
> > carries the sense of "themselves" or "each other".  For all I know
> > this is
> > fine in Klingon, but it does seem odd to me.
> >
> > Alternatively we could make a short-form sentence, where the verb
> > prefix
> > indicates the indirect object agreement:
> >
> > yuch nob
> >
> > I don't see anything to prevent this, but it's even more ambiguous
> > than the
> > long form.  Is there a way to make the "each other"/"themselves"
> > meaning
> > explicit?
> >
> > Normally -'egh and -chuq are used to indicate reflexive direct
> > objects.
> > Since they're the only reflexive markers we have, is there any way
> > to use
> > them to indicate that indirect objects are reflexives?  Or, for that
> > matter,
> > is there some other known way to unambiguously indicate reflexive
> > indirect
> > objects?
> >
> > Given the Type 1 suffixes, reflexive direct objects should be
> > straightforward - and the long form seems to be.  "They sold
> > themselves to
> > the Emperor."
> >
> > ta'vaD ngev'eghpu'
> >
> > Or, if you would like a miniature O. Henry story implied in your
> > grammar
> > examples:
> >
> > ta'vaD ngevchuqpu'
> >
> > The short form is a problem, though.
> >
> > ??? lungev'egh'pu'
> >
> > The short form requires a prefix agreeing with the direct object -
> > but the
> > rules for reflexive endings in 4.2.1 of TKD indicate that the type 1
> > endings
> > occur with the "no object" prefixes.  Which rule takes precedence?
> > Does the
> > 4.2.1 rule prevent the formation of a short-form ditransitive with a
> > reflexive direct object?  Or does the short-form rule override the
> > 4.2.1
> > rule, allowing verbs with -'egh and -chuq to take other prefixes in
> > this
> > context?
> >
> > If anyone has examples of correct usage that I can generalize from,
> > or can
> > point me to a rule I've overlooked, I would greatly appreciate it.
> > Thank
> > you!
> >
> > * Or "are not co-indexed", if you prefer.
> >
> >
> >
>
>
>
>
>

Back to archive top level