Talk:Turkic languages

Classification

• attributive attr = things that act like adjectives
• predicative pred
• substantive subst = things that act like nouns
• adverbial advl = things that act like adverbs (??)

Hierarchy

• noun (default 'subst') + DECL-NOUN

  noun->adj = n.attr + DECL-ADJ !! No <comp>arison levels though

• adj (default 'attr') + DECL-ADJ

  adj->noun = adj.subst + DECL-NOUN

• num (default 'attr') + DECL-NUM

  num->noun = num.subst + DECL-NOUN

• prn (default 'subst') + DECL-NOUN

• det (default 'attr') + NO-DECL

• v

  v->noun = v.ger + DECL-NOUN

  v->adj = v.glp + DECL-ADJ

  v->adv = v.prc + DECL-ADV

Types of non-finite verbal forms:

• Adverbial participle: <gnL> (e.g. <gnc> "Conditional adverbial participle")
• Verbal adjectives: <gpL> (e.g. <gpi> "Imperfect verbal adjective")
• Gerunds: <gerN> (e.g. <ger1> "Past/present gerund")
• Participles: <prcN> (e.g. <prc1> "Realis participle")

What about 'cop' and 'pred'

• The copula is i- (p.79)
  • -(y) (pres)
  • -(y)DI (past)
  • -(y)mIş (evid)
  • -(y)sA (cond)

Заметки разрешении морфологической неоднозначности

Arguments against just having a different tag:

 e.g. güzel<n>/güzel<adj>

• We lose the tag denoting the principle function of the stem
• We can't tell the CG to choose the principal function
• We can't tell the difference between `real' N/A ambiguity and "derivation" ambiguity

Arguments against just piling one tag ontop of another:

 e.g. güzel<adj>/güzel<adj><n>

• Having two POS in a word makes things confusing
• Having two POS tags in a word makes it difficult to write CG rules

Arguments against having a "zero derivation":

 e.g. güzel<adj>/güzel<adj><D_n><n>

• It's ugly and stupid
• Having two POS tags in a word makes it difficult to write CG rules

Прилагательное

güzel            'beautiful'          güzel<adj>/güzel<adj><subst>/güzel<adj><advl>
güzelim          'my beauty'          güzel<adj><subst><px1sg>
güzel konuştu    'she spoke well'     güzel<adj>/güzel<adj><subst>/güzel<adj><advl>
güzel bir köpek  'a beautiful dog'    güzel<adj>/güzel<adj><subst>/güzel<adj><advl>

küçük            'small'              küçük<adj>/küçük<adj><subst>/küçük<adj><advl>
küçük kızlar     'little girls'       küçük<adj>/küçük<adj><subst>/küçük<adj><advl>
küçükler         'little one(s)'      küçük<adj><subst><pl>/küçük<n>+i<cop><pres><p3><pl>

kötü             'bad'                kötü<adj>/kötü<adj><subst>/kötü<adj><advl>
kötü araba       '(a) bad car'        kötü<adj>/kötü<adj><subst>/kötü<adj><advl>
kötü yüzmek      'to swim badly'      kötü<adj>/kötü<adj><subst>/kötü<adj><advl>

Наречии

şimde            'now'                şimde<adv>
şimdelerde       'nowadays'           şimdelerde<adv>

I think you want both of these as <adv>. Historically it's something like "şu emdi<n??>" and "şu emdilerde/emdi<n??><pl><loc>", but for our purposes this is irrelevant. —Firespeaker 19:40, 26 February 2012 (UTC)

More to the point, this isn't any sort of productive process we're seeing here; my point is that it's an isolated productive-looking form because of its unique history. —Firespeaker 19:41, 26 February 2012 (UTC)

Имема существительные

evdeki          'the one in the house'     ev<n><locattr>/ev<n><locsub>
evdekinde       'in the one in the house'  ev<n><locsub><loc>

Разное

$ echo Evlerimizdeymişler | hfst-proc tr-cv.automorf.hfst 
^Evlerimizdeymişler/Ev<n><pl><px1pl><loc>+i<cop><evid><p3><pl>$

Compound tenses

Things to think about:

• analysis length:
  • ^келген эмеспи/кел<v><iv><neg><past><p3><pl>+бы<qst>/кел<v><iv><neg><past><p3><sg>+бы<qst>/кел<vaux><neg><past><p3><pl>+бы<qst>/кел<vaux><neg><past><p3><sg>+бы<qst>$, vs.
  • ^келген/кел<v><iv><past>/кел<vaux><past>$ ^эмеспи/эмес<neg><p3><sg>+бы<qst>/эмес<neg><p3><pl>+бы<qst>$
• tag/morpheme reordering should be done by transfer, such as Turkish->Chuvash negative imperative, Chuvash->Turkish possessives.
• what about different spacing, do you ever get >1 space, or nbsp or formatting between e.g. келген and эмеспи ? -- or anything that isn't a single ascii space ?

Resources

Following is a mail to the Corpora list. Might be a good idea to have a 'Resources' page/section for Turkic languages, as it is done on language pages.

Message: 6 Date: Wed, 25 Jun 2014 19:54:04 +0200 From: "Christian Chiarcos" <christian.chiarcos@web.de> Subject: Re: [Corpora-List] Turkic dictionaries To: "corpora@uib.no" <corpora@uib.no>

Dear all,

I would like to thank everyone who responded to my request and who helped me in personal conversation, in particular, Emily Bender, Jost Gippert, Max Ionov, Irina Nevskaya, Monika Rind-Pawlowski, Vit Suchomel, Francis Tyers, and Mardan Wushouer. Please find a summary, with URLs, brief description and licensing information below (no particular order):

(A) Dictionaries/Wordlists in machine-readable formats

(A.1) Gilles Sérasset's DBnary http://kaiko.getalp.org/about-dbnary/ machine-readable (RDF) dictionaries generated from Wiktionary, incl. Turkish CC-BY-SA

(A.2) Mardan Wushouer's wordlists http://www.ai.soc.i.kyoto-u.ac.jp/~mardan/resource/Chinese_Uyghur_Bilingual_Dictionary_v1.zip http://www.ai.soc.i.kyoto-u.ac.jp/~mardan/resource/Chinese_Kazakh_Bilingual_Dictionary_v1.zip http://www.ai.soc.i.kyoto-u.ac.jp/~mardan/resource/Uyghur_Kazakh_Bilingual_Dictionary_v1.zip plain word lists for Chinese-Uyghur, Chinese-Kazakh, Uyghur-Kazakh CC-BY-NC

(A.3) Altaic etymological dictionary http://starling.rinet.ru/cgi-bin/bdescr.cgi?root=config&morpho=0&basename=\data\alt\turcet includes 26 Turkic languages, available online and as DBase dump copyright restricted

(A.4) Freelang http://freelang.net English (and partially, French) word lists for 28 Turkic languages (mostly small), proprietary list format freeware (i.e., no modification)

(A.5) Apertium Turkic http://wiki.apertium.org/wiki/Turkic_languages#Pairs word lists for Turkic-Azeri, Kazakh-Tatar, 12 more pairs of Turkic languages under development open source (hosted at Sourceforge)

(A.6) RELISH http://tla.mpi.nl/relish/ lexicons for Chalkan and Tuva, provided by the RELISH project available online, XML licensing to be clarified

(A.7) PanLex http://panlex.org huge collection of word lists in a unified representation (SQL, RDF) incl. Azeri, Gagauz, Kazakh, Kirgiz, Turkish, Turkmen, Uzbek, etc. different (mostly open) licenses depending on the original source

(A.8) Intercontinental Dictionary Series http://lingweb.eva.mpg.de/ids/, http://datahub.io/de/dataset/ids word lists of minimal core vocabulary Azeri, Kumyk, Nogai, Terekeme (Azerbaijan dialect) plain text or RDF CC-BY-NC-ND

(B) Human-readable dictionaries/wordlists that can be easily converted into machine-readable formats

(B.1) Wiktionary, various languages (see A.1) http://wiktionary.org incl. Azeri, Kazakh, Kirgiz, Tatar, Turkish, Turkmen CC-BY-SA

(B.2) Chalkan dictionary http://sprachen.sprachsignale.de/tschalkanisch/tschalkanisch.html German available for academic use, with attribution, non-commercial

(B.3) Shorica http://shoriya.ngpi.rdtc.ru/ Shor dictionary and corpus copyright to be clarified, currently offline (last accessed mid-May 2014)

(B.4) Karachay-Balkar dictionary http://www.elbrusoid.org/dictionary/ Karachay-Balkar - Russian dictionary copyright restricted

(B.5) Tatar dictionary http://tatar.com.ru/dict/dict.php Tatar-Russian dictionary copyright restricted

(B.6) Khakassian dictionary http://khakas.altaica.ru/dictionary/ Khakas - English and Khakas - Russian copyright restricted

(C) other resources

(C.1) Altaica http://altaica.narod.ru/e_v-turks.htm link and resource collection, includes machine-readable and human-readable dictionaries for 17 Turkic languages (not replicated above)

(C.2) Pre-Islamic Old Turkic Texts (VATEC) http://vatec2.fkidg1.uni-frankfurt.de/ glossed corpus (XML) from which a German-Old Turkic word list can be compiled copyright restricted

(C.3) Glosbe http://glosbe.com online access to word lists and translation memories Azeri, Karachay-Balkar, Kazakh, Tatar, Turkish, Turkmen, Uzbek, etc. free online API (with severe capacity limits)

Certainly, this list is not exhaustive, so if you feel something important is missing or incorrect, please let me know ;)

All the best, Christian