Re: [Aramorph-users] XML tables

[Pierrick Brihaye]

Hi again,

Ahmed El-dawy wrote:

> Actually I made this part like this because it is so similar to this
> of the sample at LDC.

The problem is that people from the LDC know what they are talking about :-)

> Also it keeps the xml files small if we consider
> this matter.

You're right, but since we will provide the files in a jar file, the 
redundancy (and thus the compression rate) will be roughly the same.

> Anyway, it is changed now to be more readable. Actually,
> I don't know the meaning of the word (pos) till now :)

POS = "part of speech".

> > May be changed to :
> > <glosses>
> >  <gloss>and</gloss>
> >  <gloss>by/with</gloss>
> > </glosses>
>
> Yes, you are right at this. I've changed it in the new version.

Thank you. On this point, my implementation differs from the original 
one because I try to "shift" prefixes and suffixes from the stem 
definition (and may even generate words with a NO_STEM type, eg. bihi, 
fyha...). Splitting the values in the XML files makes this process more 
obvious IMHO.

We may also go further however by introducing <prefix-pos/gloss>, 
<stem-pos/gloss> and <suffix-pos/gloss> which are *very* accurate in the 
stems dictionary. What's your mind on this point ?

> > And, of course, the arabic words sould be encoded... in arabic.
>
> I will do it after making the xml files, maybe at the same program who
> translates current dictionary to xml files.

It would be a good idea.

> By the way, there's a
> problem if we transformed to xml using the transliteration. One symbol
> used is (>) which is already used for closing tag names in XML. We
> will have to transform this into &gt;.

Correct. This shouldn't be a problem if the files are generated through 
a SAX parser that will handle this escaping automatically.

> > Regarding, the stems dictionary, the format has to be slightly different
> > because we have additional information (see
> > http://www.nongnu.org/aramorph/english/dictionaries.html) :
> >
> > <root>ktb</root>
> > <lemmaID>katab-u_1</lemmaID>
> >
> > and, maybe, a "normalised" lemma
> > <lemma>katab</lemmaID>
> I know that the lemma is the one at a line starting with two
> semicolons (;;), but what is this root?

;--- ktb

We first have to check if this formalism is consistent throughout the 
stems dictionary...

> > Regarding the compatibility tables, something like this would be nice :
>
> See the current version (attached) and tell me

Since the DTDs are very short, you shoud embed then in the XML files. 
See http://www.thescarms.com/XML/DTDTutorial.asp.

Regarding :

<grammatical-categories>
  <grammatical-category>wa/CONJ</grammatical-category>
</grammatical-categories>

we may consider a :

grammatical-categories|grammatical-category

content-model. I don't know if it's accurate though since such a less 
verbose format may introduce unnecessary processing difficulties.

Well, using an XML file format would greatly help us in providing a Web 
interface that could allow adding new words in the dictionaries.

Cheers,

--
Pierrick Brihaye, informaticien
Service régional de l'Inventaire
DRAC Bretagne
mailto:address@hidden
+33 (0)2 99 29 67 78