<div dir="ltr">Thanks again Berthold! Joshua Crowgey (student working on Lushootseed)<div>has done an external FOMA-based morphological analyzer, and I think</div><div>David Inman (student working on Nuu-chah-nulth) will be doing the same, so</div><div>we probably won't be exercising the LKB morphology that way, but it's good to</div><div>have the info. As for the details of what's going on in those languages, I have</div><div>to defer to them.</div><div><br></div><div>Emily</div></div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Sep 15, 2015 at 3:25 PM, Berthold Crysmann <span dir="ltr"><<a href="mailto:berthold.crysmann@gmail.com" target="_blank">berthold.crysmann@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div text="#000000" bgcolor="#FFFFFF">
<div>Hi again, <br><span class="">
<br>
On 15/09/15 16:40, Emily M. Bender wrote:<br>
</span></div><span class="">
<blockquote type="cite">
<div dir="ltr">Thanks, Berthold, for the explanation. This may
turn out to be useful
<div>for Lushootseed and Nuu-chah-nulth, though the situation
there is somewhat</div>
</div>
</blockquote>
<blockquote type="cite">
<div dir="ltr">
<div>different. In particular, I'm intrigued that you are
treating __REDUP__ and</div>
<div>the base as separate lexical items. Is this because of the
whitespace</div>
<div>conventions of the language? Or because it was easier to
do things this</div>
<div>way than completely within the morphology? </div>
<div><br>
</div>
</div>
</blockquote>
<br></span>
Hausa does use space or hyphen to separate these guys. Remember: we
do not yet have a solution for bog-standard German compounds. So I
am lucky there: decnt language, possibly there've been some Brits
around at some point... <br>
<br>
But there is also tonal evidence suggesting that we are really
dealing with two minimal morphological words: There is no tone
copying in any morphological construiction of the language, and
there are no word other than total reduplication involving two tonal
spreading processes. Moreover, the string to be copied is easily
defined in terms of minimal word, but hard to define in terms of
managable prosodic structures (segment, syllable, foot). Hausa also
has partial reduplication (pluarctionals, some plurals), and I do
that in the morphology: just try out "sun kakkaranta littafi" (they
repeatedly read the book). <br>
<br>
How complex is the material you need to copy? Things like C1V2C3
C1V2C3 are probably best done in the morphology. Just create two
identical letter sets for the consonants, and there you are. I
am/was talking total reduplication in this thread so far (thinking
of Indonesian). <br>
<br>
E.g. for CVC reduplication:<br>
<br>
!c = ptkbdgfsxvzX<br>
!k = ptkbdgfsxvzX<br>
!v = aeiou<br>
<br>
%prefix (!c!v!k !c!v!k!c!v!k)<br>
<br>
I guess there are rules similar to this in HaG. If you are
intersted, I can chase them up. I did notice, when doing umlaut in
GG, that with increasing numbers of characters to be memorised, the
memory footprint on the LKB went up considerably. But that is years
ago, when 2GB was luxury. <br>
<br>
One specific issue: the total reduplication phrasal rule is also a
subtype of word-or-lexrule-min, so it can be a head daughter in
constructions that operate on lexical signs only (e.g. to combine
with bound demonstratives "joji jojin nan sun zo" ). I think of it
as a branching lexical rule, somehow. <br><span class="">
<br>
<blockquote type="cite">
<div dir="ltr">
<div>Also, you said (in a separate exchange we had) that the
older solution was </div>
<div>non-compositional ... presumably because you were using the
native lex entry </div>
<div>for the redup form and then squashing its EPs. <br>
</div>
</div>
</blockquote></span>
Right.<span class=""><br>
<blockquote type="cite">
<div dir="ltr">
<div> Am I understanding correctly </div>
<div>that this current solution is compositional because the
ersatz item is </div>
<div>semantically empty?</div>
</div>
</blockquote>
<br></span>
Exactly! Actually, the old analysis still works in the LKB. I use a
type addendum to impose an empty RELS list for ace and pet to get
rid of the spurious ambiguity. There's an extra file ace-types.tdl
now that is loaded by pet and ace. <br>
<br>
I have not tested much of this in Pet. It appears to work with the
three telling examples I tried just now. Shall do more systematic
testing, once I have got the time. <br>
<br>
The issue Woodley fixed related to the representation of
orthographemic changes in the AVMs: now every change is recorded on
every individual step in the derivation, as it should be, so you can
chirurgically select the point where you want to grab that
information for copying. That wasn't the case before, so you have to
*upgrade to the current svn version* imperatively! I am putting this
here to avoid fruitless experimenting on old versions of our
engines. <br>
<br>
Cheers, <br>
<br>
Berthold <br><div><div class="h5">
<br>
<blockquote type="cite">
<div dir="ltr">
<div>Emily</div>
<div><br>
</div>
</div>
<div class="gmail_extra"><br>
<div class="gmail_quote">On Sun, Sep 13, 2015 at 6:45 PM,
Francis Bond <span dir="ltr"><<a href="mailto:bond@ieee.org" target="_blank">bond@ieee.org</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Thanks
Berthold (and Woodley).<br>
<br>
This looks very promising, if a bit daunting (my mastery of
the chart<br>
mapping machinery is still shaky). We will try it out, and
probably<br>
pester<br>
you with more questions.<br>
<div>
<div><br>
On Mon, Sep 14, 2015 at 8:00 AM, Berthold Crysmann<br>
<<a href="mailto:berthold.crysmann@gmail.com" target="_blank">berthold.crysmann@gmail.com</a>>
wrote:<br>
> Hi Francis, Emily, and who else might be listening
on this,<br>
><br>
> I have now got a version of my revised analysis
working (thanks also to<br>
> Woodley for fixing a critical issue in ace for me).<br>
><br>
> Here's a brief description (for more details look
at the HaG code<br>
> $LOGONROOT/llf/hag, or ask me):<br>
><br>
> Hausa has some 4 inflectional classes that use
total reduplication. While<br>
> tonal patterns may differ between base and
reduplicant, the reduplicant does<br>
> not undergo inflection different from the base.
Thus, on the string level<br>
> (once we have dealt with suprasegmental markings),
the reduplicant is<br>
> faithful to the base (lucky for me). We do get
further inflectional marking<br>
> on the base, however.<br>
><br>
> E.g. in plural class 12, we get<br>
><br>
> nâs nâs `nurses'<br>
> sìkêt sìkêt `skirts'<br>
> jōjì jōjì `judges'<br>
><br>
> The latter can be morphologically possessed,
yielding "jōjì jōjiìnmù" (our<br>
> judges).<br>
><br>
> So in this class, the reduplicant is just a copy of
the base lexeme's orth<br>
> value (+ tone + length).<br>
><br>
> A more interesting case are augmentative
adjectives. Here the base for the<br>
> productive formation is partially reduplicated in
the masc and fem singular,<br>
> but one drops the partial reduplication in the
plural, to have total<br>
> reduplication of a clipped base instead.<br>
><br>
> E.g. class 14:<br>
><br>
> mālàm bundumēmḕ -> mā̀làmai bundumā bùndùmā̀<br>
><br>
> Which can also undergo type raising, giving<br>
><br>
> bundumā bùndùmàn mā̀làmai<br>
><br>
> Note that the inflection -n on the base does not
get copied to the<br>
> reduplicant.<br>
><br>
><br>
> As for the analysis, here is what I now do (in
ace):<br>
><br>
> 1. Parsing<br>
><br>
> As a first step, I trigger an ersatz, in token
mapping, on every item in the<br>
> chart that is followed by some other item.<br>
> The surface string I substitute is __REDUP__, which
corresponds to exactly<br>
> one lexical item (redup_n).<br>
> I tried filtering already on this first step on
partial identity of form<br>
> (with regexps), but that did not seem to work. The
ersatzing records the<br>
> original string in +CARG. The rule applies after
processing of<br>
> suprasegmentals (tone and length), and copies that
information over as well.<br>
> See tmr/redup.tdl.<br>
><br>
> In a second step, I shall use lexical filtering,
which applies after lookup<br>
> and morphological processing, to get rid of
unlicensed reduplication entries<br>
> in the chart. I'll probably implement that
beginning of next week.<br>
><br>
> During morphological processing, I introduce
constraints for the reduplicant<br>
> as part of the morphological rules applying to the
base. In particular, this<br>
> enables me to select precisely at which step in the
derivation I want to<br>
> memoise the identity of the base. All the
constraints on the reduplicant are<br>
> collected in MORPH.MCLASS.--REDUP. See rule
n_pl12_lr (irules.tdl), or, even<br>
> better, n_pl14_lr.<br>
><br>
> In syntax, I use a binary rule to combine the
reduplicant ersatz with the<br>
> base and impose all the constraints the base has
for the shape of the<br>
> reduplicant (rule n-pl-reduplication in rules.tdl).<br>
><br>
> Different plural formation patterns require
different levels of identity:<br>
> e.g. class 12 copies the segments and
suprasegmentals of the lexeme, whereas<br>
> class 14 copies the segments of the derived plural
and imposes a fixed H+<br>
> pattern on the reduplicant, and a fixed L+ tone
pattern on the base. Either<br>
> of them exempts further inflectional markings from
reduplication.<br>
><br>
> This is all done by the morphological rules
applying to the base, which<br>
> also store the constraints re the reduplicant in
MORPH.--REDUP, having a<br>
> value for orthography (--STEM string) and
suprasegmentals (--SUPRA supra).<br>
> See the rules for class 12 and class 14 in
irules.tdl. The binary<br>
> reduplication rule then imposes these constraints
on the reduplicant.<br>
><br>
> Some sample sentences for you to test:<br>
><br>
> joji jojinmu sun zo<br>
> joji jojin sun zo<br>
> malamai bunduma bunduma sun zo<br>
><br>
> In the hag directory, you can test with<br>
><br>
> ace -l -g ace/hausa.dat<br>
><br>
> for parsing<br>
><br>
> or with<br>
><br>
> ace -T -g ace/hausa.dat|ace -e -l --show-gen-chart
-g ace/hausa.g.dat<br>
><br>
> for generation.<br>
><br>
><br>
> 2. Generation<br>
><br>
> Using a single generic entry for all reduplicants,
I can now trigger the one<br>
> I need based on very general properties, i.e.
plural in Hausa. Look at<br>
> reduplication.mtr for reference. Since there's only
one rule left now, it is<br>
> soon going to move into the main trigger.mtr.<br>
><br>
> Using an ersatz, I have to replace the generic
__REDUP__ phonology with<br>
> something sensible at some point: I use
post-generation chart mapping to<br>
> copy over the relevant string from the base. The
relevant rule is the third<br>
> or fourth one down in tmr/post-generation.tdl.
Likewise, I copy over the<br>
> constraints on tone and vowel length (SUPRA) as
well, which are imposed in<br>
> the --REDUP.--SUPRA features.<br>
><br>
><br>
> 3. Conclusion<br>
><br>
> To summarise, the new solution scales up to open
classes, and it is fully<br>
> supported in ace. As for the LKB, you can test with
the ersatz as input,<br>
> e.g. "__redup__ joji sun zo" parses and generates.
Generation will only use<br>
> the ersatz. I shall look into refining my user
function to get proper<br>
> generation output.<br>
><br>
> For the moment, I kept the old analysis alive for
testing in the LKB: if you<br>
> type in "nas nas sun zo", you should still get an
analysis, and you can even<br>
> generate from it (using the ersatz). In ace, this
"native" analysis is<br>
> disabled by requiring an empty RELS on the
reduplicant (as a type addendum<br>
> on the phrasal type), which will filter out the
native contentful entry, and<br>
> keep the semantically empty one (the ersatz), thus
avoiding any spurious<br>
> ambiguity. Since the LKB analyses are a superset of
the ace analyses, there<br>
> will not be any issue w.r.t. treebanking.<br>
><br>
> As I feel about it, total reduplication is the
killer argument for having<br>
> chart mapping in the LKB, or else for having a
complete ace-based<br>
> development environment.<br>
><br>
> 4. Outlook: Chinese and Indonesian<br>
><br>
> As far as I can tell, my approach for Hausa should
be straightforward to<br>
> port to these two languages. There is the question
of the X: you would<br>
> probably treat that as a morphophonological effect
of some rule application<br>
> (I guess). So, given our technologies, the most
straightforward thing to do<br>
> is place all morphophonological changes on the
base. That leaves you free to<br>
> impose simple total identity on the reduplicant:
just choose the right<br>
> moment in the derivation (thanks to Woodley who has
just fixed the recording<br>
> of orth values in ace an hour or two ago). If that
should not be viable<br>
> (there are languages like that), one needs to
replicate some of the<br>
> morphology in token mapping. For reference, look at
how the ERG deals with<br>
> plurals of unknown words.<br>
><br>
><br>
> All the best,<br>
><br>
> Berthold<br>
><br>
><br>
> On 08/09/15 16:26, Berthold Crysmann wrote:<br>
>><br>
>> Hi Francis,<br>
>><br>
>> I guess you only worry about total
reduplication. While in principle in<br>
>> Chinese you could get away with using string
unification (thanks to the<br>
>> script, word length is limited), but specifying
the letter set will not be<br>
>> much for for either humans or machines....<br>
>><br>
>> I am currently working on total reduplication
in my Hausa grammar. I had a<br>
>> first solution that is non-compositional in the
semantics. I.e. I just use a<br>
>> binary rule that glues the stuff together,
conditioned on identity of<br>
>> predicates, but throws away the semantic
contribution in the reduplicant.<br>
>> Works in parsing, but needs *item-specific*
trigger rules in generation.<br>
>><br>
>> Right now, I am exploring with ersatzing.
Things should work well as long<br>
>> as only one of the reduplicants undergoes
additional morphology. Otherwise,<br>
>> you'll have to memoise parts of the original
string, so you can apply<br>
>> regular morphophonological changes to the
reduplicant. Seems to work in the<br>
>> LKB.<br>
>><br>
>> I shall commit that new analysis very soon. I
shall also send you a<br>
>> detailed description.<br>
>><br>
>> Cheers,<br>
>><br>
>> Berthold<br>
>><br>
>> On 08/09/15 15:28, Francis Bond wrote:<br>
>>><br>
>>> G'day,<br>
>>><br>
>>> we are working with a couple of languages
where we would like to be<br>
>>> able to write lexical rules that do things
like:<br>
>>><br>
>>> Take a two character word in Chinese (AB)<br>
>>> AB -> ABAB<br>
>>> AB -> AABB<br>
>>> AB -> AAB<br>
>>> AB -> ABXAB (where X is fixed by the
rule)<br>
>>> AB -> AXAB<br>
>>> AB -> AXB<br>
>>><br>
>>> Take a one character word (A):<br>
>>> A -> AA<br>
>>> A -> AXA<br>
>>> A -> AAX<br>
>>><br>
>>> In Indonesian we want to take an arbitrary
word and produce a duplicate<br>
>>> w -> w-w (kasus -> kasus-kasus)<br>
>>><br>
>>> More examples here:<br>
>>> <a href="http://moin.delph-in.net/LADChineseReduplication" rel="noreferrer" target="_blank">http://moin.delph-in.net/LADChineseReduplication</a><br>
>>> and<br>
>>> <a href="http://moin.delph-in.net/LADChineseAnotA" rel="noreferrer" target="_blank">http://moin.delph-in.net/LADChineseAnotA</a><br>
>>><br>
>>> Is this (or some of this) possible with the
DELPH-IN tools? If so,<br>
>>> can someone explain how to do it (or point
to a paper or website that<br>
>>> tells us how to do it)?<br>
>>><br>
>>> Thanks in advance,<br>
>>><br>
>>><br>
>><br>
>><br>
><br>
><br>
> --<br>
> Berthold Crysmann<<a href="mailto:crysmann@linguist.jussieu.fr" target="_blank">crysmann@linguist.jussieu.fr</a>><br>
> CNRS, Laboratoire de linguistique formelle (UMR
7110), U Paris Diderot<br>
> Case 7031, 5 rue Thomas Mann, 75205 Paris cedex 13<br>
> Bureau 545, bâtiment Olympe de Gouges, rue Albert
Einstein, 75013 Paris<br>
><br>
<br>
<br>
<br>
</div>
</div>
<div>
<div>--<br>
Francis Bond <<a href="http://www3.ntu.edu.sg/home/fcbond/" rel="noreferrer" target="_blank">http://www3.ntu.edu.sg/home/fcbond/</a>><br>
Division of Linguistics and Multilingual Studies<br>
Nanyang Technological University<br>
<br>
</div>
</div>
</blockquote>
</div>
<br>
<br clear="all">
<div><br>
</div>
-- <br>
<div>
<div dir="ltr">Emily M. Bender<br>
Professor, Department of Linguistics<br>
Check out CLMS on facebook! <a href="http://www.facebook.com/uwclma" target="_blank">http://www.facebook.com/uwclma</a><br>
</div>
</div>
</div>
</blockquote>
<br>
<br>
<pre cols="72">--
Berthold Crysmann <a href="mailto:crysmann@linguist.jussieu.fr" target="_blank"><crysmann@linguist.jussieu.fr></a>
CNRS, Laboratoire de linguistique formelle (UMR 7110), U Paris Diderot
Case 7031, 5 rue Thomas Mann, 75205 Paris cedex 13
Bureau 545, bâtiment Olympe de Gouges, rue Albert Einstein, 75013 Paris</pre>
</div></div></div>
</blockquote></div><br><br clear="all"><div><br></div>-- <br><div class="gmail_signature"><div dir="ltr">Emily M. Bender<br>Professor, Department of Linguistics<br>Check out CLMS on facebook! <a href="http://www.facebook.com/uwclma" target="_blank">http://www.facebook.com/uwclma</a><br></div></div>
</div>