Re: decomp(osition) module, final discussion this Friday 29th of May, 16:00 CET

Dear Philipp, All

here you can find my observations on the decomp module. I am sorry to raise
problems rather than providing solutions. However, I am not criticizing the
model, while suggesting improvements in its description.

--------

<cite>
Decomposition is the process of indicating which elements constitute a
multi-word or compound lexical entry
</cite>

What is the difference between a multi-word and compound lexical entry? The
word "compound" only appears in the decomposition module.

--------

<cite>
The simplest way to do this is by means of subterm property, which
indicates that a lexical entry is a part of another entry.
</cite>

I fear that this sentence is misleading, since it could suggest that the
lexical entry at the subject position is a part of the lexical entry at the
object position, while the right direction is exactly the opposite one.

--------

In Example decomp/example1, the lexical entry :AmphibiousLanding is related
only to the lexical entry :Landing, i.e. :Amphibious is not reported as a
subterm. Is it motivated by some convention for the compilation of lexica,
or is it simply motivated by the fact that the current lexicon does not
contain an entry for :Amphibious? Or is it just an omission?

Additionally, do we want to also provide the canonical form of each entry?
Actually, the example could suggest that the decomposition module avoids
the need for the canonical form.

--------

<cite>
The use of the property subterm has two limitations. First, we can not
indicate inflectional properties of the lexical entry when appearing as a
subterm of another term
</cite>

What do you mean by "inflectional property"

--------

I think that the specification should spend more effort (a couple of
sentences) to better explain what a component is supposed to be. Why can't
I simply use lexical entries? According to my understanding of the decomp
module, the reason is that certain information is not bound to the lexical
entry itself, but to the specific use of a lexical entry as a constituent
of a decomposition. I think that this is the reason why "we can not
indicate inflectional properties of the lexical entry when appearing as a
subterm of another term", and if I am right, we should explain it better in
example2.

--------

<cite>
A component is declared as a subclass of rdf:sequence
</cite>

The class should be rdf:Seq.

--------

In the definition of Component:

"The class component represents a part of a lexical entry, that is either a
word, sequence of words, phrase, or affix"

Is it the component or the lexical entry to be either a word, sequence of
words, phrase or affix?

--------

The property correspondsTo does not appear in the diagram. I suspect that
it replaced identifies.

--------

In Example decomp/example2, I think it would be better to continue the
example on :AmphibiousLanding, showing how the use of components allows
overcoming the limitations of subterms. Moreover, I would describe in prose
the meaning of the statements using the predicates genders and number.

--------

The Example decomp/example3 shows the use of rdf:_N to indicate the order
of constituents. However, this example is more elaborated, since it also
shows how a constituent can be further decomposed into its constituents. So
I would suggest to introduce an example 2B, which only deals with the
position of constituents, just before example3, which also shows the
decomposition of constituents.

--------

I think that the description of Example decomp/example4 should be
substantially extended. As far as I understand, the purpose is to represent
via the module decomp how the syntactic frame is realized in text.

--------

Philipp said he had not updated some ontologies, thus I didn't look either
at ontologies or at examples. However, I noticed that images are not
aligned to the wiki, maybe because they have not been updated or the
backing example in the GIT repository is no update, either.

--------

Best regards

Manuel Fiorelli


2015-05-26 21:51 GMT+02:00 Philipp Cimiano <cimiano@cit-ec.uni-bielefeld.de>
:

> Dear all,
>
>  I open the decomp module for comments. The module is quite concise, so I
> do not expect major issues.
>
> The final discussion of the decomp module is scheduled for this Friday
> 16:00 (CET).
>
> Please send any comments or issues to me until Thursday.
>
> Kind regards,
>
> Philipp.
>
> --
> --
> Prof. Dr. Philipp Cimiano
> AG Semantic Computing
> Exzellenzcluster für Cognitive Interaction Technology (CITEC)
> Universität Bielefeld
>
> Tel: +49 521 106 12249
> Fax: +49 521 106 6560
> Mail: cimiano@cit-ec.uni-bielefeld.de
>
> Office CITEC-2.307
> Universitätsstr. 21-25
> 33615 Bielefeld, NRW
> Germany
>
>
>


-- 
Manuel Fiorelli