universaldependencies / ud_arabic-padt Goto Github PK

For example:

# sent_id = ummah.20040531.0026:p3u2
# text = دهعملاب راشتسملا دمحأ ىفطصم روتكدلا كلذ نلعأ.
...
8       دهعملا  دهَعمَ    NOUN    N------S2D      Case=Gen|Definite=Def|Number=Sing       6       nmod    6:nmod:بِ:gen    Vform=دِهَعمَلاَ|Gloss=institute|Root=`_h_d|Translit=al-maʿhadi|LTranslit=maʿhad
9       .       .       PUNCT   G---------      _       1       punct   1:punct Vform=.|Translit=.

This could very well be an ignorance of Arabic, since I know almost nothing about the language. However, I would expect either line 8 to have SpaceAfter=No in this sentence, or for the text to have a space before the final period.

If this is indeed a data issue, there are many such sentences in PADT. I can volunteer to send a list of all of them, or to edit the file if you tell me which of the two changes are needed. Thanks!

Hi,
I think there is a minor format error in the MISC field for words which contain with 'alif maddah (like in آن).
The transliteration of 'alif maddah uses the vertical bar, which should only be used as a field separator.
e.g.

sent_id = afp.20000815.0006:p3u1:
150 آن آن X U--------- _ 151 nmod 151:nmod Vform=آن|Root=OOV|Translit=|n

sent_id = ummah.20040715.0013:p7u1
47 الكلدآشوريين الكلدآشوريين X U--------- _ 42 nmod 42:nmod Vform=الكلدآشوريين|Root=OOV|Translit=Alkld|$wryyn

sent_id = afp.20000815.0110:p5u1
45 آحادي آحادي X U--------- _ 47 nmod 47:nmod Vform=آحادي|Root=OOV|Translit=|HAdy

May be you could use "ā" or "â" instead ? It is a very minor problem (32 cases in ar_padt-ud-train.conllu, 9 in test and just 1 in dev)
Thanks,
regards
Johannes

From UniversalDependencies/docs#836:

Arabic hua, though it is attached to the predication as obl for some reason, in PADT?

Quite likely a conversion error. PADT has been converted using a procedure originally developed for Czech, and the two languages differ substantially in copular constructions.

There are several cases where text has a period in the middle of the sentence. One of my Arabic speaking labmates identified those sentences as headlines. On account of that, it may very well be intentional to not split them, but it is somewhat confusing.

# newpar id = xinhua.20030509.0162:p4
# sent_id = xinhua.20030509.0162:p4u1
# text = واضاف " ان كل من يتصور انه يعرف المدة التى سيستغرقها هذا الامر ، انما يخدع نفسه . انه امر غير معروف ".

# newdoc id = alhayat.20011206.0109
# newpar id = alhayat.20011206.0109:p1
# sent_id = alhayat.20011206.0109:p1u1
# text = الانتهاء من المواصفات القياسية للسلع المتبادلة الاسبوع المقبل . الصادرات المصرية الى العراق 628،1 بليون دولار في 01 شهور

# newdoc id = alhayat.20011117.0097
# newpar id = alhayat.20011117.0097:p1
# sent_id = alhayat.20011117.0097:p1u1
# text = وقعت اتفاقاً للتعاون المالي مع المانيا بقيمة 227 مليون مارك . مصر تتوقع انخفاضاً في حجم المعونات

There are plenty of others I can dig up. Is this intended? Thanks!