Reduplication with Hermit-Crab parser -- possibly bugs?

[sebadr...@gmail.com]

Dear all,

 

I am trying to implement reduplication, using the phonological-rule-based parser (HermitCrab.NET).

 

In Awetí, there is only full reduplication of verb stems.  Active verbs that end in a consonant, when reduplicated, are of a certain inflection class (“E.V.”, for “E-Verb”, which show a final thematic vowel -e), to which vowel-final stems (reduplicated or not) never belong.

 

To account for the two variants (vowel-final and consonant-final), I believe I will have to use allomorph-processes.

 

As the change of inflection classes is accounted for independently of the allomorph-processes, I have created two entries, both as abstract forms so that the parser does not get confused by the lexeme form. 

 

In one entry, (“lll”, for testing purposes), I have the process:

   X [C] --> 1 2 + 1 2, and here I indicate the inflection class change to E.V.

 

The allomorph process in the other entry (“ll”, for testing purposes) is also:

   X [V] --> 1 2 + 1 2, and here I indicate the default inflection class (“N.EA.V”) as result

 

The first time I tried it, the parsing when using "Try a word" was slowed down too much (more than 5 minutes even for simple cases without reduplication).  As in similar situations, I deleted the entries and now I am trying to create them again.

 

Now parsing generally works at a reasonable speed, but parsing of reduplication fails. 

In the following sequence of screenshots, I show here only the correct path of parsing of the simple form atoto, with morphs a- (1sg) and to (‘to go’).  For some reason, when  synthesizing the correct analysis, the process stops before adding the (empty) suffix (occupying the obligatory slot in the case of the default inflection class which is occupied by -e in the case of the E-Verbs) and the person prefix.

 

 

The same happens for a consonant-final active verb such as ‘to sit down’ with the stem tik , which should go into the special inflection class E.V. marked by a thematic vowel -e at the end of the stem in the reduplicated form atiktige (a- tik + tik -e, with lenition of the final k).

Again, the synthesis stops after having applied the reduplication process, without even trying to apply any further affixation, which it correctly had identified in the preceding analysis part:

 

 

There is also all kind of strange parsing output which does not seem to be warranted by the reduplication rules at all:

 

 

I suspect that these may be bugs in the HermitCrab parser, but I may be wrong.  If so, what can I do to fix this, if anything?

 

As always, thanks in advance,

Sebastian

[Michael Maxwell]

Sebastian, I'm trying to understand a couple things.

First, your test for the vowel-final case ends in a consonant. So I
wouldn't expect the rule, which calls for a word-final vowel, to apply.
Am I missing something?

Second, this "inflection class *change*" (emphasis added): inflectional
rules don't change the inflection class, and certainly allomorphs of
inflectional affixes don't change the inflection class. Are you saying
your test lexical entries have each been assigned the respective
inflection classes *in the dictionary*?

Mike Maxwell

On 5/21/2024 3:17 PM, sebadr...@gmail.com wrote:
> Dear all,
>
> I am trying to implement reduplication, using the
> phonological-rule-based parser (HermitCrab.NET).
>
> In Awetí, there is only full reduplication of verb stems.  Active verbs
> that end in a consonant, when reduplicated, are of a certain inflection

> class (“E.V.”, for “E-Verb”, which show a final thematic vowel */-e/*),

> to which vowel-final stems (reduplicated or not) never belong.
>
> To account for the two variants (vowel-final and consonant-final), I
> believe I will have to use allomorph-processes.
>
> As the change of inflection classes is accounted for independently of
> the allomorph-processes, I have created two entries, both as abstract
> forms so that the parser does not get confused by the lexeme form.
>

> In one entry, (“*lll*”, for testing purposes), I have the process:

>
>    X [C] --> 1 2 + 1 2, and here I indicate the inflection class change
> to E.V.
>

> The allomorph process in the other entry (“*ll*”, for testing purposes)

> is also:
>
>    X [V] --> 1 2 + 1 2, and here I indicate the default inflection
> class (“N.EA.V”) as result
>
> The first time I tried it, the parsing when using "Try a word" was
> slowed down too much (more than 5 minutes even for simple cases without
> reduplication).  As in similar situations, I deleted the entries and now
> I am trying to create them again.
>
> Now parsing generally works at a reasonable speed, but parsing of
> reduplication fails.
>
> In the following sequence of screenshots, I show here only the correct

> path of parsing of the simple form */atoto/*, with morphs */a-/* (1sg)
> and */to/* (‘to go’).  For some reason, when  synthesizing the correct

> analysis, the process stops before adding the (empty) suffix (occupying
> the obligatory slot in the case of the default inflection class which is

> occupied by */-e/* in the case of the E-Verbs) and the person prefix.

>
> The same happens for a consonant-final active verb such as ‘to sit down’

> with the stem */tik/* , which should go into the special inflection
> class E.V. marked by a thematic vowel */-e/* at the end of the stem in
> the reduplicated form */atiktige/* (*/a- tik + tik -e/*, with lenition
> of the final */k/*).

>
> Again, the synthesis stops after having applied the reduplication
> process, without even trying to apply any further affixation, which it
> correctly had identified in the preceding analysis part:
>

> There is also all kind of strange parsing output which does not seem to
> be warranted by the reduplication rules at all:
>

> I suspect that these may be bugs in the HermitCrab parser, but I may be
> wrong.  If so, what can I do to fix this, if anything?
>
> As always, thanks in advance,
>
> Sebastian
>

> --
> "FLEx list" messages are public. Only members can post.
> flex_d...@sil.org
> http://groups.google.com/group/flex-list
> <http://groups.google.com/group/flex-list>.
> ---
> You received this message because you are subscribed to the Google
> Groups "FLEx list" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to flex-list+...@googlegroups.com
> <mailto:flex-list+...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/flex-list/005901daabb3%2482304c70%248690e550%24%40gmail.com <https://groups.google.com/d/msgid/flex-list/005901daabb3%2482304c70%248690e550%24%40gmail.com?utm_medium=email&utm_source=footer>.

[Andy Black]

Hi, Sebastian.

I think the problem is with the derivational reduplication suffixes mapping to an active verb (av) and then expecting that an intransitive verb (vi) template will be found when vi is under av in the category hierarchy.  It does not work that direction.  It *is* the case that any template for the higher av will be tried for the lower vi stem, but not the other way around.

I added a sense to lll that maps from av to vi.  Then it worked:



--Andy

--

"FLEx list" messages are public. Only members can post.
flex_d...@sil.org

http://groups.google.com/group/flex-list.

---
You received this message because you are subscribed to the Google Groups "FLEx list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to flex-list+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/flex-list/005901daabb3%2482304c70%248690e550%24%40gmail.com.

[Sebastian Drude]

Thanks a lot to both.  I will check and answer in my afternoon (in some 4 hours). Sebastian

[Sebastian Drude]

Dear Mike,

here my answers.

First, your test for the vowel-final case ends in a consonant. So I wouldn't expect the rule, which calls for a word-final vowel, to apply. Am I missing something?

I am not sure where the misunderstanding lies in this case... the stem of the vowel-final verb is to, the reduplicated 1sg. form is atoto.

 

Second, this "inflection class *change*" (emphasis added): inflectional
rules don't change the inflection class, and certainly allomorphs of
inflectional affixes don't change the inflection class. Are you saying
your test lexical entries have each been assigned the respective
inflection classes *in the dictionary*?

Reduplication in Awetí is a DERIVATIONAL process; I should made have clearer.  And yes, my stems are tagged as E-Verbs if they belong to this class.  For instance, the stem tik is the stem of an E-Verb 'to sit'.  The first singular (unmarked for aspect etc.) form is atige.  The corresponding reduplicated form is atiktige; only the stem is reduplicated, and then the thematic vowel is added.  Also the reduplicated verbs from regular (Non-e-)Verbs with consonant-final stems belong to the E-Verb class: tan 'to run', 1sg. atan; reduplicated 1sg. atantane.

Verbs which do not belong to the E-Verb-class are not tagged; the non-E-Verb-class is the default inflection class.

As to my problems, it seems Andy solved them, but anyways, I wanted to clarify these points, and to thank you for caring to bother and answer.

Sebastian

 

[Sebastian Drude]

Thanks again, Andy!
I tried this, not with an av>vi sense, but with two senses, vi>vi and vt>vt.  Then it worked here as well.

Now I understand the implication hierarchy of rules and templates better.

Thanks a lot again!

Sebastian

Em terça-feira, 21 de maio de 2024 às 18:49:38 UTC-3, Andy Black escreveu: