Turkic languages

From Apertium
Revision as of 13:31, 16 June 2014 by Ilnar.salimzyan (talk | contribs) (→‎Official poem: manual language identification with probability 90%)
Jump to navigation Jump to search

The Turkic languages include Turkish, Azeri, Uzbek, Uyghur, Kazakh, Kyrgyz, Tatar, Chuvash, Turkmen, Bashkir, Sakha, and several dozen other languages. The languages are related with varying levels of mutual intelligibility. A team called Apertium Turkic aims to enhance MT coverage of these languages with apertium.

The ultimate master plan involves generating independent finite-state transducers for each language, and then making individual dictionaries and transfer rules for every pair. The current status of these goals is listed below.

Status

The ultimate goal is to have multi-purpose transducers for a variety of Turkic languages. These can then be paired for X→Y translation with the addition of a CG for language X and transfer rules / dictionary for the pair X→Y. Below is listed development progress for each language's transducers and dictionary pairs.

Transducers

See also: Turkic lexicon and Languages

Once a transducer has ~80% coverage on a range of medium-large corpora we can say it is "working". Over 90% and it can be considered to be "production".

name Language native name ISO 639 formalism state stems coverage location primary authors
-2 -3
apertium-kaz Kazakh қазақ тілі kk kaz HFST (lexc+twol) production 36,595 ~94.5% apertium-kaz (languages) Ilnar, Jonathan, Fran, Nathan
apertium-tat Tatar татар теле tt tat HFST (lexc+twol) production 55,702 ~91% apertium-tat (languages) Ilnar, Fran, Jonathan, Röstäm
apertium-kir Kyrgyz кыргыз тили ky kir HFST (lexc+twol) working 14,424 ~90.4% apertium-kir (languages) Jonathan, Mirlan, Fran
apertium-tur Turkish Türk dili tr tur HFST (lexc+twol) working 17,221 ~87.3% apertium-tur (incubator) Fran, Gianluca
apertium-chv Chuvash чӑваш чӗлхи cv chv HFST (lexc+twol) development 8,579 ~85% apertium-chv (languages) Hèctor
apertium-kum Kumyk къумукъ тил - kum HFST (lexc+twol) development 4,918 ~90.2% apertium-kum (languages) Fran, Jonathan
apertium-uzb Uzbek o'zbek tili uz uzb HFST (lexc+twol) development 34,470 ~82.9% apertium-uzb (languages)
apertium-bak Bashkir башҡорт теле ba bak HFST (lexc+twol) development 2,827 ~66% apertium-bak (languages) Fran, Jonathan, Ilnar, Milli
apertium-nog Nogay Ногай тили - nog HFST (lexc+twol) development 1,385 ~81.4% apertium-nog (languages) Fran, Jonathan
apertium-tuk Turkmen Türkmençe tk tuk HFST (lexc+twol) development 2,988 ~70.7% apertium-tuk (languages) Fran
apertium-kaa Karakalpak Qaraqalpaq tili - kaa HFST (lexc+twol) prototype 25,545 ~86.1% apertium-kaa (incubator) Beknazar, Fran, Jonathan
apertium-uig Uyghur ئۇيغۇر تىلى ug uig HFST (lexc+twol) prototype 17,585 ~54.2% apertium-uig (incubator) Jonathan, Märdan, Fran
apertium-crh Crimean Tatar Qırımtatar tili - crh HFST (lexc+twol) prototype 11,757 ~85.4% apertium-crh (incubator) Fran
apertium-sah Yakut Саха тыла - sah HFST (lexc+twol) prototype 11,531 ~89.6% apertium-sah (incubator) Fran
apertium-tyv Tuvan Тыва дыл - tyv HFST (lexc+twol) prototype 11,695 ~92.7% apertium-tyv (incubator) Fran
apertium-kjh Khakas Хакас тілі - kjh HFST (lexc+twol) prototype 293 ~39.3% apertium-kjh (incubator) Fran
apertium-aze Azerbaijani Azərbaycan dili az aze SFST not known to work apertium-tur-aze (staging) Gianluca

Turkic languages by subgroup

Pairs

Some Turkic languages that are particularly similar to one another (and hence have high levels of mutual intelligibility) include those in the following list:

Chuvash is very distant from other Turkic languages and is not even partially mutually intelligible with any of them.

Table of existing pairs

See also: Turkic-Turkic translator

Text in italics denotes language pairs in the incubator. Regular text denotes a developing language pair in nursery, while text in bold denotes a stable well-working language pair in trunk and text in bold and italics denotes a pair in staging. Bidix stems as counted with dixcounter are displayed below.

tur aze tuk uzb kir kaz tat bak chv nog kaa kum uig
tur - tur-aze
tuk-tur
tur-uzb
tur-kir
tur-tat
cv-tr
aze tur-aze
-
tuk tuk-tur
-
uzb tur-uzb
- kir-uzb
kir tur-kir
kir-uzb
- kaz-kir
tat-kir
kaz kaz-kir
- 'kaz-tat
'
nog-kaz
kaz-kaa
kaz-kum
kaz-uig
tat tur-tat
tat-kir
'kaz-tat
'
- tat-bak
cv-tt
bak tat-bak
-
chv cv-tr
cv-tt
-
nog nog-kaz
-
kaa kaz-kaa
-
kum kaz-kum
-
uig kaz-uig
-

Other pairs

tur aze tuk uzb kir kaz tat bak chv nog kaa kum uig
eng tr-en
ky-en
eng-kaz
epo eo-tr
khk khk-kaz
rus tat-rus
cv-ru

Roadmap

  • Stable release of apertium-kaz
  • Stable release of apertium-tat
  • Stable release of apertium-kaz-tat
  • Rework apertium-kir to match new standards
  • Bring apertium-bak up to date (based on apertium-tat)
  • Expand apertium-tat-bak
  • Beta release of apertium-kaz-kir
  • Beta release of apertium-tur-kir
  • Expand apertium-tuk
  • Expand apertium-chv
  • Basic transducers for:

Getting involved

We have a work plan for developing Turkic-Turkic translators and are working on a how-to for building a Turkic lexicon. Please come talk to us on IRC or contact us on the apertium-turkic mailing list.

Samples

Article 1 of the Universal Declaration of Human Rights:

All human beings are born free and equal in dignity and rights. They are endowed with reason and conscience and should act towards one another in a spirit of brotherhood.

Language Text
Kazakh Барлық адамдар тумысынан азат және қадір‐қасиеті мен кұқықтары тең болып дүниеге келеді. Адамдарға ақыл‐парасат, ар‐ождан берілген, сондықтан олар бір‐бірімен туыстық, бауырмалдық қарым‐қатынас жасаулары тиіс.
Tatar Барлык кешеләр дә азат һәм үз абруйлары һәм хокуклары ягыннан тиң булып туалар. Аларга акыл һәм вөҗдан бирелгән һәм бер-берсенә карата туганнарча мөнәсәбәттә булырга тиешләр.
Kirghiz Бардык адамдар өз беделинде жана укуктарында эркин жана тең укуктуу болуп жаралат. Алардын аң‐сезими менен абийири бар жана бири‐бирине бир туугандык мамилекылууга тийиш.
Turkish Bütün insanlar hür, haysiyet ve haklar bakımından eşit doğarlar. Akıl ve vicdana sahiptirler ve birbirlerine karşı kardeşlik zihniyeti ile hareket etmelidirler.
Uzbek (Latin) Barcha odamlar erkin, qadr-qimmat ba huquqlarda tang bo'lib tug'iladilar. Ular aql va vijdon sohibidilar va bir-birla iga birodarlarcha muomala qilishlari zarur.
Tuva Бүгү кижилер хостуг база мөзүзү болгаш эргелери дең кылдыр төрүттүнер. Оларга угаансарыыл болгаш арын-нүүр бердинген болур болгаш олар бот-боттарынга акы-дуңмалышкы хамаарылганы көргүзер ужурлуг.

Vulnerability

This table summarizes the vulnerability of various Turkic languages. Vulnerability data is derived from the ‘Atlas of the World’s Languages in Danger, © UNESCO, http://www.unesco.org/culture/languages-atlas’ and Ethnologue.

Language ISO639-3 Location Speakers Status
Ethnologue UNESCO
Chagatai chg Turkmenistan 0 10 (Extinct) -
Salchuq slq Iran 0 10 (Extinct) -
Chulym clw Russian Federation 44 8b (Nearly extinct) 4 (Critically endangered)
Karaim kdr Lithuania & Ukraine 56 8b (Nearly extinct) 5 (Extinct)
Tofa (Karagas) kim Russian Federation 90 8b (Nearly extinct) 4 (Critically endangered)
Krimchak jct Ukraine 200 8b (Nearly extinct) -
Ili Turki ili China 120 8a (Moribund) 3 (Severely endangered)
Yugur (Western/Saryg) ybe China 4,600 8a (Moribund) 3 (Severely endangered)
Balkan Gagauz Turkish bgx Turkey 331,000 7 (Shifting) -
Dolgan dlg Russian Federation 1,050 6b (Threatened) 2 (Definitely endangered)
Shor cjs Russian Federation 2,840 6b (Threatened) 3 (Severely endangered)
Ainu aib China 6,570 6b (Threatened) -
Altay, Northern atv Russian Federation 57,400 6b (Threatened) 3 (Severely endangered)
Salar slr China 70,000 6b (Threatened) 1 (Vulnerable)
Urum uum Georgia, Russian Federation, Ukraine 192,700 6b (Threatened) 2 (Definitely endangered)
Khalaj klj Iran 42,100 6a (Vigorous) 1 (Vulnerable)
Khorasani kmz Iran 400,000 6a (Vigorous) 1 (Vulnerable)
Kashkay qxq Iran 1,500,000 6a (Vigorous) -
Khakas kjh Russian Federation 42,610 5 (Developing) 5 (Extinct)
Nogay nog Russian Federation 87,410 5 (Developing) 2 (Definitely endangered)
Gagauz gag Moldova 173,920 5 (Developing) -
Karakalpak kaa Uzbekistan 410,410 5 (Developing) -
Kumyk kum Russian Federation 426,550 5 (Developing) 1 (Vulnerable)
Tuva tyv China, Mongolia, Russian Federation 283,400 4 (Educational) 1 (Vulnerable)
Karachay-Balkar krc Russian Federation 310,730 4 (Educational) 1 (Vulnerable)
Crimean Tatar crh Ukraine 475,540 4 (Educational) -
Chuvash chv Russian Federation 1,077,420 4 (Educational) 1 (Vulnerable)
Bashkort bak Russian Federation 1,221,340 4 (Educational) 1 (Vulnerable)
Uzbek, Southern uzs Afghanistan 2,961,980 3 (Wider communication) -
Azerbaijani, Southern azb Iran 16,913,280 3 (Wider communication) -
Altay, Southern alt Russian Federation 57,400 2 (Provincial) 2 (Definitely endangered)
Yakut sah Russian Federation 450,000 2 (Provincial) 1 (Vulnerable)
Tatar tat Russian Federation 5,407,550 2 (Provincial) 2 (Definitely endangered)
Uyghur uig China 8,791,610 2 (Provincial) -
Kyrgyz kir Kyrgyzstan 2,941,930 1 (National) -
Azerbaijani, Northern azj Azerbaijan 7,324,270 1 (National) -
Turkmen tuk Turkmenistan 7,560,560 1 (National) -
Kazakh kaz Kazakhstan 8,077,770 1 (National) -
Uzbek, Northern uzn Uzbekistan 18,968,250 1 (National) -
Turkish tur Turkey 50,733,420 1 (National) -

Official poem

  • (tur) Kovayla bira içerim, ama sen bilmezsin. Yarın gelir misin?
  • (aze) Vedrəyle pivə içirəm, ama sen bilməzsən. Yarın gələrmisən?
  • (tat) Чиләк белән сыра эчәм, әмма син белмисең. Иртәгә киләсеңме?
  • (kaz) Шелекпен сыра ішемін, бірақ сен білмейсің. Ертең келесің бе?
  • (kir) Чака менен сыра ичем, бирок сен билбейсиң. Эртең келесиңби?
  • Челек булан пиво ичемен амма сен билмейсен. Эртен гелемисен?
  • Chelak bilan pivo ichaman lekin sen bilmaysan. Ertaga kelasanmi?
  • Shelek penen pivo ishemen, biraq sen bilmeysen'. Erten' kelesen'be?

Footnotes


Bibliography

  • Washington, J. N., Salimzyanov, I., and Tyers, F. M. (2014) "Finite-state morphological transducers for three Kypchak languages". Proceedings of the 9th Conference on Language Resources and Evaluation, LREC2014
  • Salimzyanov, I., Washington, J. N., and Tyers, F. M. (2013) "A free/open-source Kazakh-Tatar machine translation system". Proceedings of MT Summit XIV
  • Washington, J. N., Ipasov, M. and Tyers, F. M. (2012) "A finite-state morphological analyser for Kyrgyz". Proceedings of the 8th Conference on Language Resources and Evaluation, LREC2012
  • Tyers, F. M., Washington, J. N., Salimzyanov, I., and Batalov, R. (2012) "A prototype machine translation system for Tatar and Bashkir based on free/open-source components". Proceedings of the Turkic Languages Workshop at the Language Resources and Evaluation Conference, LREC2012