Mova0.6 pre-release

the flag of Kazakhstan 1. Standard Kazakh and its varieties

Standard Kazakh is the shared norm used in education, public writing and formal communication in Kazakhstan. Kazakh speakers also use regional forms, informal speech and the languages of the communities in which they live. These kinds of variation can overlap. The kk edition describes the standard used in Kazakhstan and identifies other usages where they affect how a text or utterance is understood.

The standard used in Kazakhstan

Kazakh is the state language of Kazakhstan. The contemporary standard supplies the spelling and grammatical conventions used in Kazakh-medium education, public administration, textbooks, edited news and other public writing. It also has spoken uses, including formal announcements, prepared broadcasts and classroom presentations. A speaker who uses the standard in those settings can use local vocabulary, informal forms and a regional pronunciation in everyday conversation.

Cyrillic is the established script of this standard in current schooling and public writing. The Kazakh Cyrillic alphabet has letters for distinctions that matter in Kazakh, including қ and к, ғ and г, and ң and н. A change of script can preserve these distinctions if it represents the same words and grammatical endings. Script therefore gives information about the written form and its likely readership. It takes evidence from the words and their use to establish a regional or national usage.

Two research resources illustrate different uses of standard written Kazakh. UD Kazakh KTB supplies texts analyzed in Universal Dependencies, a framework for representing the syntactic relations between words. The Kazakh portion of FLORES-200 identified as kaz_Cyrl supplies Cyrillic text for evaluating translation. The treebank documents grammatical analyses of its texts, and the evaluation set documents a written variety in a translation setting. Their scope is narrower than the range of settings in which people speak Kazakh (Universal Dependencies contributors 2025, § Summary; NLLB Team 2022, § kaz_Cyrl).

An utterance can be short, personal and ordinary while following the standard. The following examples use the same spelling and inflectional conventions in a statement and a question. In the glosses, PL is plural, DAT dative, PST past, NPST nonpast, 1PL first person plural, 1SG first person singular, 3 third person and Q a question particle. Hyphens in a gloss show morphological boundaries.

FormRomanizationGlossTranslation
Балалар мектепке барды.Balalar mektepke bardı.child-PL school-DAT go-PST-3The children went to school. (constructed)
Ертең барамыз ба?Erteñ baramız ba?tomorrow go-NPST-1PL QAre we going tomorrow? (constructed)

Within Kazakhstan, Russian is also used in public life and is officially used alongside Kazakh in state bodies and local self-government. Individual speakers and institutions differ in how they distribute the two languages across settings. In a given institution, the language of a public document, the language of an internal discussion and the language used to address a particular visitor can each follow a different convention.

Regional variation within Kazakhstan

Regional variation within Kazakhstan is most readily heard in conversation. It can involve the choice or meaning of a word, the pronunciation of a sound or sequence, and the frequency of a particular expression. Everyday vocabulary such as names for clothing, foods, tools and forms of address often shows local preferences. Many such preferences are shared across nearby settlements, and a single settlement can contain speakers with different family and migration histories.

Broad descriptions use labels such as western, southern and northeastern to locate overlapping concentrations of features. The labels cover sizeable areas, and the distribution of one form can cross the area associated with another. Regional speech shares the main resources of Kazakh grammar, including case marking, possession and verbal inflection. Features of that common system, such as the order of constituents in a basic clause, can be compared with those of other languages using the Kazakh entry and linked feature descriptions in WALS (Dryer and Haspelmath 2013, § Kazakh).

Lexical variation can affect interpretation even when the syntax is familiar. For example, сым sım has the widespread sense ‘wire’ and a regional clothing sense ‘trousers’. The standard word for the clothing is шалбар šalbar. The verb кию ‘put on, wear’ supports the clothing sense in the first example below. In a conversation about electrical repairs, a different context would support the wire sense.

FormRomanizationGlossTranslation
Ол сым киді.Ol sım kidi.3SG trousers put.on-PST-3They put on trousers. (constructed)
Ол шалбар киді.Ol šalbar kidi.3SG trousers put.on-PST-3They put on trousers. (constructed)

A regional form can also become familiar well beyond the places where it is most frequent. Speakers moving for education, work or family bring local expressions into new networks. A form used mainly at home can appear in a recorded interview, dialogue in fiction or a message between friends. Its occurrence in writing can reflect the setting and the speaker’s voice. Its distribution across speakers and contexts is what supports a regional attribution.

Kazakh across national borders

Kazakh-speaking communities extend across the borders of Kazakhstan. They include communities in China, Mongolia, Russia and Uzbekistan. Some communities have longstanding local roots and others include more recent migrants. National borders affect schooling, publishing and public language use, so the Kazakh learned and used in one community can differ from that used in another.

In China, Kazakh is used especially in Xinjiang. A modified Arabic script has an established role in Kazakh writing there. Local schooling, publishing and contact with Chinese shape the vocabulary speakers encounter, particularly for institutions and public life. Speakers can also encounter the Cyrillic standard through material from Kazakhstan. Those who learn both scripts can recognize shared Kazakh words and endings across the two forms of writing.

In Mongolia, Kazakh is especially associated with Bayan-Ölgii and is also used by communities elsewhere in the country. Cyrillic is established in Kazakh writing there. Kazakh-medium schooling and contact with Mongolian are both relevant to literacy and language use. The extent to which a speaker uses Mongolian in education, public life or work depends on the setting and the individual’s experience.

In Russia and Uzbekistan, Kazakh is used in communities with varying access to Kazakh-language schooling and publications. Russian and Uzbek, respectively, are important contact languages. Cyrillic Kazakh materials can circulate in these communities, while the scripts and conventions a reader uses in the surrounding language depend on that reader’s schooling and local institutions. Public vocabulary may differ where countries organize and name institutions differently.

These differences can produce stable local conventions in schools, publications or broadcasting. A national variety refers to such conventions and patterns of use within a country. A regional variety can extend across a border, and several regional varieties can be used in the same country. Family speech, schooling and the publications a speaker reads are therefore relevant when identifying a particular usage. The model for spelling and formal grammar in this edition remains the standard used in Kazakhstan.

Script and schooling across communities

Schooling links a written norm to a community. It gives pupils regular exposure to spellings, grammatical terminology and genres such as essays, reports and official correspondence. A pupil who initially learns to read and write Kazakh in Cyrillic and one who initially learns it in Arabic script can know many of the same spoken forms. Their reading fluency and familiarity with spelling conventions can differ. In multilingual schools and communities, pupils may also acquire literacy in a second language, sometimes using a different script.

SettingWriting conventions relevant to Kazakh
KazakhstanCyrillic is established in public writing and schooling. Latin spellings also occur in romanization and in material associated with alphabet reform.
Kazakh communities in ChinaModified Arabic script has an established role in Kazakh texts. Reading Cyrillic Kazakh depends on a reader’s separate exposure to it.
Kazakh communities in MongoliaCyrillic is established in Kazakh texts. Readers may also use Cyrillic for Mongolian, which has its own spelling conventions.
Kazakh communities in Russia and UzbekistanCyrillic Kazakh materials are used. Access to Kazakh-medium instruction and familiarity with other writing conventions vary by community.

The same familiar word can illustrate the distinction between script and vocabulary. The Arabic-script form in the table is read from right to left. The romanization here serves to read the examples.

FormRomanizationGlossTranslation
қазақqazaqKazakhKazakh (constructed)
قازاقqazaqKazakhKazakh (constructed)

Kazakh also has a tradition of Arabic-script educational writing associated with Akhmet Baitursynuly. His 1914 grammar is an example of Kazakh used to teach and describe Kazakh in that script. It belongs to the history of Kazakh schooling and literacy. Current conventions in a community are established by the materials used there today (Baitursynuly 1914, title page).

Latin spellings have different purposes. A romanization may help someone read a name, supply a rendering in a document written in another language or let a user type Kazakh with an available keyboard. An alphabet proposal is intended to supply a broader set of spelling conventions. The interpretation of a particular Latin spelling depends on the convention in use. Even when two written forms can be read as the same Kazakh word, converting a whole text requires identifying the spelling conventions of both scripts.

Contact and variation in spoken Kazakh

Speech varies with audience, topic and setting as well as place. Pronunciation and endings may be reduced in familiar conversation, especially at a faster pace. In informal speech, for example, a first-person nonpast form in -мын or -мін can occur with a shorter ending. The full form is usual in standard edited writing. A transcription or dialogue representing informal speech can preserve the shorter form. Both examples below have the same subject and time reference.

FormRomanizationGlossTranslation
Ертең барамын.Erteñ baramın.tomorrow go-NPST-1SGI’ll go tomorrow. (constructed)
Ертең барам.Erteñ baram.tomorrow go-NPST-1SGI’ll go tomorrow. (constructed)

Sustained contact with other languages can affect vocabulary and pronunciation. An established borrowing can occur in an otherwise Kazakh sentence and take ordinary Kazakh inflection. In вагонға ‘into the train car’, the noun вагон ‘train car’, borrowed through Russian, takes the Kazakh dative -ға. In bilingual conversation, speakers can also switch languages within a turn, sometimes at a clause boundary. In the second example, a Kazakh main clause is followed by a Russian conditional clause.

FormRomanizationGlossTranslation
Вагонға отырдық.Vagonğa otırdıq.train.car-DAT sit-PST-1PLWe got into the train car. (constructed)
Мен ертең келемін, если будет время.Men erteñ kelemin, esli budet vremya.1SG tomorrow come-NPST-1SG if be.FUT.3SG timeI’ll come tomorrow if there’s time. (constructed)

A borrowed word can remain in use among speakers who rarely use the language it came from. Code-switching depends on the speakers’ shared repertoire and the immediate conversation. Speakers may switch to quote someone, address a different participant or use an expression familiar to their interlocutor. In a written transcript, the surrounding words and their grammatical relations help establish whether a stretch is part of a Kazakh clause or a switch to another language.

Diglossia describes a relatively stable division of social functions between varieties or languages. To apply it to a particular community, identify which variety or language is conventionally used in which settings. The use of edited standard Kazakh in school writing and local Kazakh at home is one possible division within Kazakh. Conventional choices between Kazakh and Russian, Chinese, Mongolian or Uzbek can also distribute functions across languages. The details depend on the community, institution and generation. A bilingual utterance by itself shows what a speaker did on that occasion. Evidence of regular choices across settings is needed to describe a stable division of functions.

How Mova handles regional and national variation

The first level of tools/mova-parse uses tools/langseg to assign a language to each sentence and span. Its Latin-script recognition covers English and French. It also recognizes Hebrew-script varieties and scholarly transliterations of Hebrew or Aramaic. Scripts outside its coverage are assigned other. Kazakh Cyrillic and the modified Arabic script used for Kazakh are outside that coverage. Latin-script Kazakh has no recognition entry among the supported Latin-script languages.

At the second level, tools/mova-parse uses each span’s language model to parse it. Kazakh has no model for kk. For a span with a missing model, tools/mova-parse records the condition in the # parse comment, leaves LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL as _, and sets Parse=None in MISC. Kazakh has no tokenization, tagging, lemmatization or parsing support in Mova. The prag and world components read English trees, and there are no Kazakh trees for them to read.

Mova holds no annotated Kazakh corpus and has no measured Kazakh treebank baseline. The baseline defined for a language uses en xpos-fill before en train trains the tree parser and tagger on that language’s Universal Dependencies training split. en parse-eval measures unlabeled and labeled attachment scores on the development split.

Squid reads each grammatical point in this manual as a claim. Formalizable claims enter its queue for implementation. The rules and distinctions in this chapter are unimplemented. The first priorities are the points needed for a tokenizer, a tagger and a Universal Dependencies treebank baseline.

Sources for this chapter

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/kk/01-varieties/