Feature datasets for the Atlas are based on descriptive literature.

To create a dataset you will need the following things:

Start by creating a folder for your feature in the archive, and upload your dataset there, even if it is still a work in progress: this makes it easier to discuss any problems or questions you might have.

You can see an example of a completed feature dataset here.

For a quick reference on how to collect data, see the Step-by-step instructions. The instructions below go into more detail.

List of columns

Some of the columns are required (without them, there may be problems with rendering maps and chapters) and some are not.

Note: if you use an apostrophe (for example, in the name of a village (Tad-Magitl') or in transcription), use U+0027 '. If you are working in Excel, when loading the table to the drive, make sure that the apostrophes remain correct.

  • id - unique number for each row = one observation in your dataset
  • family - language family
  • group - language branch/group
  • lang - language name

  • idiom - the name of a language variety; descriptive sources provide information on specific dialects or village varieties of languages, and we aim to be as precise as possible about where our information comes from; this also means that if a source says: “in dialect X, the same form is used”, you can add this as an observation to your dataset
  • type - specifies whether the idiom is a village variety, a dialect spoken in multiple villages, or a standard language; please use our standard names and type classifications

In our literature database you can find bibliographical information about the source, as well as information on which idioms are represented in the sources (on the sheet source type and idiom, columns idiom and type). Alternatively, you can search for the name and type of a certain idiom in the East Caucasian villages dataset.

  • genlang_point - for our maps showing one datapoint per language we need to choose which datapoint (in case we have multiple points per language) is representative and will thus be showed on this map. For this purpose you need to set the appropriate row in this column to yes (no more than one per language), and set all other observations for the same language to no; see also Philosophy - Current approach. The full list of languages can be found here
  • map - in some cases there are multiple observations for one idiom (e.g. one frequent suffix and more marginal one to express the same meaning). On the map we can show only one value per datapoint, so you can use this column to set information you want to see on the map to yes and everything else to no.

One value = one row in the dataset. If you have multiple values for one idiom, create two rows for this idiom, each with a unique id. The general rule of thumb for the genlang_point column is 1 “yes” per language (the Dargwa varieties listed in Language sample each count as a separate language), for map it is 1 “yes” per idiom.

  • feature - the name of your feature / chapter
  • value1 - the relevant values of a specific parameter (e.g. attested / not attested). Try not to use many values for the parameter. The more values there are, the more difficult it is to perceive the final map
  • value1_name - the name of the parameter: the feature Evidentiality in the tense system, for example, has two parameters: Evidentiality as a meaning of the perfect and Evidentiality in the tense system. The name of each parameter should be capitalized. The content of value1_name will be used as the title for your map

Maps are generated based on the value1 column. If you want to show the distribution of multiple parameters, please name further columns value2, value3, etc., and accordingly: value1_name, value2_name.

  • source - reference to the source you consulted (see Literature references below)
  • page - relevant page in the source
  • comment - in this column you can “pour out your soul” in the words of G.A. Moroz: add any kind of thought on an observation or the source it appears in. Keep in mind, however, that anything you write here will be visible to users of the Atlas. So please write your comments in English, and make them clear and informative
  • contributor - your name in English, so we know how to properly credit your work

  • form - what the phoneme / morpheme / construction / lexeme looks like in Caucasiologist transcription
  • example_as_in_source - original transcription of the example
  • example - example of how the form is used (if applicable) in Caucasiologist transcription with morpheme boundaries
  • translation_as_in_source - original translation of the example
  • translation - your own English translation of the example in case the source was in Russian or you dislike the original translation for some reason
  • gloss - glosses for the example. Please follow the Leipzig glossing rules where possible. This includes unifying glosses from the original sources if necessary. Format the non-lexical glosses in CAPS.
  • example_source - reference to the source of the example
  • example_page - page reference for the example
  • example_comment - any kind of comment you would like to add regarding the example

  • date - the date on which you submitted your table; edits of the table after its first publication on the website will receive a new date stamp accordingly

Literature references

When you add a reference to your dataset, check if it is already listed in the literature database. If yes, copy the bibtexkey of the reference from the database to the source column in your table.

A bibtexkey is a unique identifier for a source which allows us to easily cite sources across the Atlas.

(You can find our library of descriptive sources here.)

In case you used multiple sources for one row / observation, separate the keys with a semicolon and space (; ), and do the same with the page numbers in the adjacent column.

If you refer to multiple page ranges from one source, separate them with a comma.

If the entire source was relevant, for example because it was a paper devoted to your topic, or because you read the whole grammar and the feature is not mentioned anywhere, indicate NA in the page column.

source page
khalilova2009; khalilova2011 221, 234–239; NA

In the example above, pages 221 and 234-239 from khalilova2009 have relevant information about the feature, while the paper khalilova2011 was consulted/is relevant in its entirety.

If the source of the information is a personal conversation, it is necessary to indicate the name of the source with a note p.c. for example Santa Claus, p.c..

Adding new literature

If you use a source which is not in our database yet, you will have to add it by submitting a form with the necessary information.

Russian resources are listed in Cyrillic script with a translation in English of the title and booktitle fields.

The bibtexkey for a source is constructed as follows:

surname (Latin script, lower case letters) + year

If there are two authors, use both surnames, cf. Chumakina & Corbett (2008) becomes chumakinacorbett2008. For sources with more than two authors, write down only the surname of the first author followed by et al, e.g. alekseevetal2008.

In case a different resource by the same author and from the same year is already present in the literature database, add a keyword following the year, e.g. bokarev1949 / bokarev1949avar.

For unpublished sources use the surname followed by the word draft, and the year in which the manuscript was produced or when it was expected to be published, for example `creisselsdraft2020’.

The ‘author’ field should be filled as follows: last name, name and (if present) the first letter of the patronymic or second name, for example: Абдуллаев, Сайгид Н. If the source has more than one author use and (in English): Абдуллаева, Айшат З. and Гаджиахмедов, Нурмагомед Э. and Кадыраджиев, Калсын С. etc. If the source does not specify the full names, try to find them on the Internet rather than using initials.

For city names in the ‘address’ field use English names, e.g. Moscow, not Москва.

Don’t forget to upload a copy of the source to the appropriate folder in the library using the bibtexkey as filename.


Some general principles for transcription:

  • we mostly follow IPA conventions but with a few notable exceptions:
    • ts = c, ʃ = š, tʃ = č, tɬ, tɬ’ = ƛ , ƛ’, dʒ = ǯ , ʒ = ž
    • nasalization is indicated with a tilde (~) above the vowel: aⁿ > ã
    • except diphtongs, because they would require double tildes, which looks pretty afwul: ũ̃̃o (in the source code these are two vowels with a tilde above it side-by-side)
    • gemination as well as long vowels are indicated with a lengthmark ː (triangles, not dots): чч > čː for consonants, ō > oː for vowels
  • here is a table with all of the phonemes attested in the languages of our sample, transcribed in IPA (column phoneme); how to transcribe them for the atlas (column TALD), and how they are represented in different idioms and sources (all other columns). If you click on a column, you can view the bibtexkey of the source (in addition to the language and idiom). This also means that if you have the bibtexkey of a grammar, you can quickly find the corresponding phoneme and transcription.


For any general questions about data collection or the library, you can always contact Chiara.

If you have a more specific question about your feature, please also contact Chiara, and she will connect you with an expert consultant for your feature.