uralex_documentation.md

UraLex 2.0 Uralic basic vocabulary with cognate and loanword information

Mervi de Heer, Mikko Heikkilä, Kaj Syrjänen, Jyri Lehtinen, Outi Vesakoski, Toni Suutari, Michael Dunn, Urho Määttä & Unni-Päivä Leino

How to cite

Please cite UraLex 2.0 as following when referring to this dataset:

De Heer, Mervi; Blokland, Rogier; Dunn, Michael; Vesakoski, Outi. (submitted manuscript). “Loanwords in basic vocabulary as an indicator of borrowing profiles.”

and

Syrjänen, Kaj; Maurits, Luke; Leino, Unni; Honkola, Terhi; Rota, Jadranka & Vesakoski, Outi. (submitted manuscript). “Crouching TIGER, Hidden Structure: Exploring the nature of linguistic data using TIGER values.”

Please include the DOI provided for UraLex 2.0 in this release in your citation as well.

Overview

The UraLex 2.0 basic vocabulary dataset has its origins in the basic vocabulary cognacy dataset collected by the research initiative BEDLAN (Biological Evolution and the Diversification of Languages), funded by the Kone Foundation between 2009-2013. The data has since been revised and expanded in follow-up research projects, including SumuraSyyni (2014-2016), UraLex (2014-2016) and AikaSyyni (2017-2020). The dataset has been compiled especially for the purposes of quantitative language classification/historical linguistics, such as Bayesian Inference of phylogeny.

The bulk of the data was originally compiled by Jyri Lehtinen, the principal compiler of the BEDLAN research initiative's cognacy dataset. It has since been expanded by the UraLex project, with the bulk of the editing work done by Kaj Syrjänen and Jyri Lehtinen. A number of individuals have made contributions to the dataset at its various stages, including Hilkka Ahola, Zoltán Balogh, Rogier Blokland, Natalia Chinaeva, Robert Forkel, Nikolett F. Gulyás, Mikko Heikkilä, Katri Hiovain, Terhi Honkola, Sulev Iva, Santra Jantunen, Eino Koponen, Svetlana Lumme, Luke Maurits, Eve Mikone, Arto Moisio, Larisa Ponomareva, Michael Rießler, Tapani Salminen, Merja Salo, Olga Titova, Marja Torikka, Judit Varga, Susanna Virtanen, Kaarina Vuolab-Lohi and Evgenia Zhivotova.

The dataset mainly covers lexical reflexes (e.g. words and expressions) denoting 313 meanings. Most of these meanings (226) come from standardized basic vocabulary lists, which cover meanings that are fairly universal, culturally neutral, and generally expressed by semantically and morphologically simple words which are relatively stable over time. Basic vocabulary words also tend to be resistant to being replaced by borrowing or semantic shift (McMahon & McMahon 2005). The 226 basic vocabulary meanings of the dataset cover the contents of three standardized basic vocabulary lists: the Swadesh200 list (Swadesh 1952), the Swadesh100 list (Swadesh 1955), and the Leipzig-Jakarta list (Tadmor 2009). The 87 meanings that are not in the 226 basic vocabulary meanings cover WOLD401-500, a list of 'less basic vocabulary' first introduced in Lehtinen et al. (2014).

The lexical reflexes representing each meaning in the dataset have been chosen to be as semantically neutral and context-independent as possible; in other words, they are contemporary general-purpose words or expressions for each meaning. Simple words or expressions have been favoured over complex ones. A language may also have multiple lexical reflexes documented for a single meaning, generally in cases where there are several equally appropriate expressions to denote that meaning.

To ensure consistent quality of the dataset, one individual has been responsible for drawing up the list of representative words for each language, and this list for most of the languages has then been double-checked by one or more language experts or native speakers. The compilers and language-checkers are listed in the Language_compilers table of the dataset.

In addition to the recorded gloss, the phonetic transcription in the Uralic Phonetic alphabet and/or International Phonetic Alphabet is included for some of the lexical reflexes.

Each lexical reflex is also associated with multistate characters that reflect what are often referred to as cognate relationships - but should more accurately be called root-meaning relationships (Chang et al. 2015). Root-meaning relationships indicate that the neutral words (i.e. words whose use is not bound to a specific context) occupying a given meaning slot originate from the same ancestral root, initially acquired through inheritance. The words are considered to belong to the same root-meaning set even if they represent different derivations, as long as they share the same historical root and occupy the same meaning slot. For instance, reflexes of the meaning 'animal' in several Saami (North, Inari, Skolt) and Finnic languages (Finnish, Karelian, Ingrian, Western Votic, Estonian and Võro) are based on the same ancestral root, the Proto-Uralic stem *elä- 'to live' (SSA, 102), using different derivational suffixes, and are consequently grouped to the same 'root-meaning' set. In contrast, cognate relationships (in the accurate sense of the word) require the full word, rather than just the root, to share a historical connection through inheritance. Cognates also do not need to occupy the same meaning slot, unlike root-meaning relationships. For example, Komi Zyryan bi 'fire', Finnish päivä 'day' and North Saami beaivvás,, beaivi 'sun, day' are considered to be cognate words that originate from a shared Proto-Finno-Ugric form, possibly the Proto-Uralic stem *päjwä, *päjwa 'warm, heat, fire' (SSA, 457). However, they do not form a common root-meaning trait because they do not occupy the same meaning slot. Likewise, the aforementioned reflexes for 'animal', which count as root-meaning forms, would not be counted as cognates, as they are different derivations of a shared root rather than full words with an ancestral connection.

In addition to inheritance from an ancestral language, linguistic items are also transmitted between languages through borrowing. While borrowed words share a root with their source language, they are not considered to belong to the same root-meaning set as their source, even if the donor language belongs to the same language family, since root-meaning sets represent inherited connections rather than lateral connections. For example, the words in the meaning slot for 'left' in Finnic languages (e.g. Est. kura 'left') are considered to be of common origin, while the North Saami gurut is a borrowing from Finnic defining a different set (EES 193). Similarly borrowings from other language families are separated into their own root-meaning sets. However, the dataset also includes another multistate character column besides the root-meaning sets to reflect correlate relationships; these group together words of the same meaning in different languages that share a root-form through either borrowing or inheritance.

For UraLex 2.0, the borrowing information drafted in the first UraLex release has been revised and updated by Mikko Heikkilä and Mervi de Heer. Heikkilä's primary focus was on the Finnic and Saamic languages. De Heer has updated and compiled loanword information for all languages in the dataset. UraLex 2.0 is published in conjunction with a detailed analysis on the contents and dynamics of borrowings in the basic vocabularies of six languages featured in the dataset (de Heer et al. 2021).

The loanword information concerns the columns borr_source, indicating the most likely donour language(s) for the borrowed lexemes, borr_qual, marking the certainty and credibility of the loan etymology, and ref_borr, containing the reference for the borrowing information. Additional notes on the loanwords can be found in the etym_notes column, which gives details on both inherited and loan etymologies when necessary. Mikko Heikkilä provided the reconstructed forms given for some items.

Other commentary on the items is placed into the general_notes column. UraLex 2.0 removes the columns regarding the age of the lexemes due to weak attestation of their absolute ages in the literature.

The sources for the borrowed lexemes given in the borr_source column were collected from etymological literature. A substantial number of borrowings in Uralex 2.0 were borrowed from ancestral stages of the donour languages into ancestral stages of the Uralic recipient languages (usually referred to as "Proto-X"). They are labelled without the "proto" denominator to eliminate redundancy unless necessary for distinguishing labels. For example, Germanic borrowings in Standard Estonian refer to Proto-Germanic loanwords acquired into Proto-Finnic. Both the older borrowings as well as loanwords from modern languages are assigned into appropriate root-meaning sets as described above. The labels for the donour languages used in Uralex 2.0 are explained in detail in de Heer et al. 2021 (see section 2.3).

A certainty estimate (possible, probable or clear) for each borrowing is marked in the borr_qual column. The estimates were given according to evaluations, critical commentary and certainty tags in literature (see de Heer et. al. 2021 for details). Clear refers to uncontroversial and transparent borrowings; for probable items the literature expresses uncertainty but no compelling alternative etymologies are given; possible borrowings still need a thorough evaluation and they have most uncertainty in literature. Information on some newer loanwords from e.g. from Russian is often not included in the limited pool of literature available for smaller Uralic languages and therefore expert evaluation was used to collect them. Etymologist Mikko Heikkilä identified the certainty estimation and the borrowing source for such loanwords (see ref_borr column). In these cases, the clear status indicates that the loanword is etymologically transparent. Probable means that there are minor difficulties in the sound correspondences between the donor and recipient languages although the borrowing is semantically plausible. No possible borrowings were identified in expert assessment.

The literature cited in the ref_borr column was selected according to guidelines presented in Junttila (2015), which identify suitable literature types for compiling Uralic loanword information. Following this, we used three main types of literature, which are primary research literature on loan etymologies, etymological dictionaries and evaluative etymological literature. During the data collection process, literature dedicated to the specific UraLex languages was prioritized. However, for those languages which do not have their own reliable etymological dictionaries or articles, information was sourced from literature focusing on other languages as well, e.g. the Finnish and Estonian etymological dictionaries (SSA, EES) were used to obtain borrowing information for smaller Finnic languages. In addition, general dictionaries with loanword information were used when applicable. For instance, the dictionary of Mari (Moisio & Saarinen 2008) pools together views on Mari lexemes of various researchers and also contains loanword information presented by them. This information is used in conjunction with literature dedicated to specific loanword layers.

We have considered literature where the authors have applied established methodology of comparative historical linguistics, i.e. the study of sounds law and sound substitution. Works that are not based on the methodology and have been clearly rejected by the scientific community (e.g. Gordeev, 1979-1983) were not taken into account. Only those etymological works which have contributed to the borrowing information in basic vocabulary are given in the references, which are not to be taken as a comprehensive list of Uralic etymology.

The root-meaning relationships and correlate relationships in the current version of the dataset are based on published etymological references, mostly those published before 2014. The loanword information is compiled from literature published by the end of 2020.

The data currently covers 26 Uralic languages: Finnish, Ingrian, Karelian, Estonian, South Estonian (Võro), Veps, (Western) Votic, (Courland) Livonian, Inari Saami, Kildin Saami, North Saami, Pite Saami, Skolt Saami, South Saami, Ume Saami, Erzya, Meadow Mari, Komi-Zyrian, Komi-Permyak, Udmurt, Hungarian, (Sosva) Mansi, (Vakh-Vasyugan) Khanty, Tundra Nenets, Nganasan and Northern Selkup. The data also includes a partial reconstruction of Proto-Uralic.

The dataset also provides several subsets of the included meanings (meaning lists). These include the lists that serve as the basis for the chosen meanings - that is, Swadesh100 (Swadesh 1955), Swadesh200 (Swadesh 1952), Leipzig-Jakarta (Tadmor 2009) and WOLD401-500 (a list of less basic vocabulary from Lehtinen et al. 2014). In addition, the dataset also includes a Swadesh207 list (which combines the meanings of Swadesh200 and Swadesh100), the Fullbasic list (which combines the 226 basic vocabulary meanings from Leipzig-Jakarta, Swadesh100 and Swadesh200 as one list) and Ura100 (a Uralic basic vocabulary list introduced in Syrjänen et al. 2013). Notably, the Ura100 list, which is based on an older 17-language version of the Uralic dataset, should be regarded as obsolete, and we do not encourage people to use it. In addition to the aforementioned meaning lists, the dataset also includes Leipzig-Jakarta ranks for the meanings that belong to the Leipzig-Jakarta list and the WOLD401-500 list.

Information from the UraLex basic vocabulary dataset or its predecessor, the BEDLAN cognate database, has featured in Honkola et al. (2013), Syrjänen et al. (2013), Lehtinen et al. (2014), List et al. (2017) and Tambets et al. (2018).

The dataset is released under Creative Commons Attribution 4.0. The authors welcome all contributions for future versions.

Contents

The raw version of the dataset (found in the raw folder of the repository) is organized into eight tables, provided as separate TSV (tab-separated values) files. The overall data structure of each table in the basic vocabulary dataset is described below, with separate sections for each table of the dataset. Fields in parentheses represent information that has been duplicated from another table of the dataset using the VLOOKUP function (see the ODS file) to make the contents more human-readable.

Citable bibliographical references from the Citation_codes and Borrowing_references tables are also provided as a separate BibTeX file. The dataset folder also includes maintenance scripts for update purposes.

In addition to the raw version the repository also includes a CLDF conversion of the data produced by Robert Forkel and Luke Maurits.

Data

This table contains the main data of the basic vocabulary dataset, including each recorded lexeme along with their correlate and root-meaning sets.

1. (language)

   Language name.

2. (definition)

   Verbose definition of a meaning.

3. (uralex_mng)

   Meaning name in UraLex/LexDB database form.

4. mng_item

   BEDLAN dataset numerical meaning code.

5. lgid3

   BEDLAN dataset numerical language code.

6. item

   Lexeme data. Contains a lexeme or [No equivalent] (no suitable equivalent for a meaning exists), [Form not found] (no suitable equivalent was found) or [Not reconstructable] (non-recontructable meanings in Proto-Uralic).

7. item_UPA

   Phonetic transcription in Uralic Phonetic Alphabet (included for 11 languages).

8. item_IPA

   Phonetic transcription in International Phonetic Alphabet (included for 16 languages).

9. form_set

   Correlate set (historical connection based on borrowing or cognacy), marked with positive integers. For [No equivalent] items the field is marked with '0'; for [Form not found] and [Not reconstructable] items the field is marked with '?'.

10. cogn_set

    Cognate set (historical connection based on cognacy), marked with one-letter or two-letter codes. For [No equivalent] items the field is marked with '0'; for [Form not found] and [Not reconstructable] items the field is marked with '?'.

11. borr_source

    Borrowing source of lexeme.

12. borr_qual

    Likelihood of borrowing (possible, probable or clear).

13. etym_notes

    Notes related to etymology of the lexeme.

14. ref_borr

Bibliographical references related to borrowed lexemes. Multiple references separated by comma and space. The references are listed in order of importance.

15. glossing_notes

    Notes related to the meaning of the lexeme.

16. general_notes

    Other notes related to the lexeme.

17. ref_abbr

    Bibliographical references related to lexeme. Multiple references separated by comma and space.

Meanings

This table provides information related to separate meanings, including the meaning codes used to refer to separate meanings, and verbose descriptions of each meaning.

1. mng_item

   BEDLAN dataset numerical meaning code.

2. LJ_rank

   Leipzig-Jakarta rank, included for meanings belonging to either WOLD401-500 or Leipzig-Jakarta and marked "-" for the remaining meanings.

3. uralex_mng

   Meaning name in UraLex/LexDB database form.

4. definition

   Verbose definition of a meaning.

Languages

This table provides language-related information. Notably, it also includes codes and information for Uralic languages that are currently not covered by the main contents of the dataset.

1. lgid3

   BEDLAN dataset numerical language code.

2. language

   Language name in verbose form.

3. ASCII_name

   Language name in simplified form.

4. ISO-639-3

   ISO-639-3 code for the language (if available).

5. Description

   Additional language description.

6. Subgroup

   Traditional subgroup of the language.

Meaning_lists

This table specifies the meaning lists and related information. The meanings found on each of the meaning lists is marked with boolean values 0 (absent) and 1 (present).

1. mng_item

   BEDLAN dataset numerical meaning code.

2. (uralex_mng)

   Meaning name in UraLex/LexDB database form.

3. (LJ_rank)

   Leipzig-Jakarta rank, included for meanings belonging to either WOLD401-500 or Leipzig-Jakarta and marked "-" for the remaining meanings.

4. Ura100

   Ura100 list (see Syrjänen et al. 2013).

5. Swadesh100

   Swadesh100 list.

6. Swadesh207

   Swadesh200 + Swadesh100 list.

7. Leipzig-Jakarta

   Leipzig-Jakarta list. Notably, the UraLex version of the list covers 101 meanings instead of 100. This is because "foot" and "leg" in UraLex are represented as separate meanings, in the style of the Swadesh200 list, and not as a combined meaning "foot/leg", as they are on the official Leipzig-Jakarta list.

8. WOLD401-500

   WOLD401-500 list (see Lehtinen et al. 2014).

9. Fullbasic

   Swadesh200 + Swadesh100 + Leipzig-Jakarta list.

10. Swadesh200

    Swadesh200 list.

Meaning_list_descriptions

This table provides more verbose descriptions for the meaning lists found in the Meaning_lists table.

1. list

   Meaning list name.

2. description

   Meaning list description.

Borrowing_references

1. abbreviation

   Bibliographical reference key for the borrowing information.

2. original_reference

   Bibliographical reference information related to key. Entries are recorded in plain text. BibTeX entries are provided as a separate BibTeX file (Borrowing_citations.bib).

3. bibtex_reference

   Keys of the BibTex entries given in the BibTex file Borrowing_citations.bib.

Citation_codes

1. ref_abbr

   Bibliographical reference key.

2. original_reference

   Bibliographical reference information related to key. Citable references (publication, URL) are recorded as BibTeX entries, whereas experts are recorded in plain text. The BibTeX entries recorded here are also provided as a separate BibTeX file (Citations.bib).

3. type

   Reference type (P = publication, U = URL, E = expert).

Language_compilers

This table records the word list collectors and double-checkers of each language.

1. lgid3

   BEDLAN dataset numerical language code.

2. (language)

   Language name.

3. collected_by

   Who has collected the data for a language.

4. final_wordlist_checked_by

   Who has checked the wordlist for a language.

References

Chang, W., C. Cathcart, D. Hall & A. Garrett. (2015). "Ancestry-constrained phylogenetic analysis supports the Indo-European steppe hypothesis." Language 91(1): 194-244. DOI: 10.1353/lan.2015.0005.

De Heer, M., Blokland, R., Dunn, M., Vesakoski, O. "Loanwords in basic vocabulary as an indicator of borrowing profiles". MS.

EES = Metsmägi, I., M. Sedrik & S. Soosaar. (2012). "Eesti etümoloogiasõnaraamat." Tallinn: Eesti Keele Sihtasutus.

Gordeev, F. I. (1979-1983), Etimologicheskiy slovar' mariyskogo yazyka. I. A-B (1979), II. V-D (1983). Mariyskoye knizhnoye izdatel'stvo, Yoshkar-Ola.

Honkola, T., O. Vesakoski, K. Korhonen, J. Lehtinen, K. Syrjänen & Niklas Wahlberg. (2013). "Cultural and climatic changes shape the evolutionary history of the Uralic languages." Journal of Evolutionary Biology 26: 1244-1253. DOI: 10.1111/jeb.12107.

Junttila, S. (2015). Tiedon kumuloituminen ja trendit lainasanatutkimuksessa: Kantasuomen balttilaislainojen tutkimushistoria. Helsinki: Helsingin yliopisto.

Lehtinen, J., T. Honkola, K. Korhonen, K. Syrjänen, N. Wahlberg, & O. Vesakoski. (2014). "Behind Family Trees: Secondary Connections in Uralic Language Networks." Language Dynamics and Change 4: 189-221. DOI: 10.1163/22105832-00402007.

List, J.-M., S. J. Greenhill & R. D. Gray. (2017). "The potential of automatic word comparison for historical linguistics." PLOS One. DOI: 10.1371/journal.pone/0170046.

McMahon, A. & R. McMahon (2005). Language Classification by Numbers. Oxford: Oxford University Press.

SSA = Suomen sanojen alkuperä I-III (1992-2000). Suomalaisen Kirjallisuuden Seuran Toimituksia 556. Kotimaisten kielten tutkimuskeskuksen julkaisuja 62. Suomalaisen Kirjallisuuden Seura, Helsinki.

Swadesh, M. (1952). "Lexicostatistic dating of prehistoric ethnic contacts." Proceedings of the American Philological Society 96, 452-463.

Swadesh, M. (1955). "Towards greater accuracy in lexicostatistic dating." International Journal of American Linguistics 21, 121-137.

Syrjänen, K., T. Honkola, K. Korhonen, J. Lehtinen, O. Vesakoski, & N. Wahlberg. (2013). "Shedding more light on language classification using basic vocabularies and phylogenetic methods. A case study of Uralic." Diachronica 30(3), 323-352. DOI: 10.1075/dia.30.3.02syr.

Tadmor, U. (2009). "Loanwords in the world's languages: Findings and results." In M. Haspelmath & U. Tadmor (eds.), Loanwords in the world's languages: A comparative handbook, 55-75. Berlin: Walter de Gruyter.

Tambets, K., B. Yunusbayev, G. Hudjashov, A.-M. Ilumäe, S. Rootsi, T. Honkola, O. Vesakoski, Q. Atkinson, P. Skoglund, A. Kushniarevich, S. Litvinov, M. Reidla, E. Metspalu, L. Saag, T. Rantanen, M. Karmin, J. Parik, S. I. Zhadanov, M. Gubina, L. D. Damba, M. Bermisheva, T. Reisberg, K. Dibirova, I. Evseeva, M. Nelis, J. Klovins, A. Metspalu, T. Esko, O. Balanovsky, E. Balanovska, E. K. Khusnutdinova, L. P. Osipova, M. Voevoda, R. Villems, T. Kivisild & M. Metspalu. (2018). "Genes reveal traces of common recent demographic history for most of the Uralic-speaking populations." Genome Biology 19(139). DOI: 10.1186/s13059-018-1522-1.