Endangered language documentation and endangered language revitalisation have been two hot topics in recent years. For instance, the United Nations Educational, Scientific and Cultural Organization (UNESCO) declared the year 2019 as the International Year of Indigenous Languages. However, although the UNESCO and many other organizations (e.g. The Endangered Languages Documentation Programme or SIL International) urge the public to be aware of the rapidly decreasing number of languages in the world, it does not slow down the annual rate of language loss. For example, the total number of speakers of the Kusunda language, a moribund language spoken in Nepal, decreased to only one person in 2020.

Introduction

I started collecting articles and lexical materials on the Kusunda language in 2017 with the hope to find time studying the origin of its speakers and its prehistoric contacts with neighboring languages. With the help of fellow scholars, I accumulated a total of 27 theses and articles. The earliest reference dates back to 1848 (Hodgson 1848). Unfortunately, I found only a handful of studies providing data on 100 or more lexical items. Subsequently, I converted Kusunda lexical material from five sources (Reinhard 1970; Rana 2002; Watters 2005; Donohue 2013; Aaley, 2017) into the formats recommended by the Cross-Linguistic Data Formats (CLDF, Forkel et al. 2018) initiative. In doing so, I met four major difficulties:

1. Data storing and sharing:
   Despite the fact that the idea of “open science” has been advocated for over a decade, “open data” is still an exception rather than the norm in the field of linguistics. Therefore, secondary linguistic data are either rare or fragmentary. Furthermore, “open” linguistic data are not persistently stored online.

2. Digitisation:
   Linguistic data accumulated over years but the majority of linguistic data is not available in digital form. This raises the problem that linguists cannot inspect the published data efficiently, and it also prevents linguists from applying further analysis on the data.

3. Standardisation:
   In Table 1, I provide three nouns and two verbs that frequently occur in linguistic studies: ‘cloud’, ‘hand’, ‘fog’, ‘to speak (1st SG)’ and ‘to move (1st SG)’. By just presenting five lexical items, one can already see wide variation in the existing Kusunda materials.

1. 1. 1. The phonological rendering of the words is not standardised according to International Phonetic Symbols (IPA). First, IPA has existed for more than 100 years, however, it is not being used regularly in existing historical linguistic data sets. Most of the time, we have encountered data recorded by customised phonetic symbols and accompanied by a guideline. For example, as Reinhard mentioned in his article, “j” and “ny” correspond to IPA /dz/ and  /nʲ/. However, it is not always the case that a guideline is provided.
      2. The basic vocabulary entries (Swadesh 1952), like ‘hand’ and ‘cloud’, have entirely different forms across the studies. Even though in the five studies conducted between 2002 and 2019 the lexical datasets were elicited from the same two speakers, Ms. Gyani Maiya and Ms. Kamala, the lexical material is highly diverse across the wordlists. Are these words the same words but in different “word forms” or do they represent the synonyms for the same concepts?
      3. It commonly happens that linguists provide a list of words with English annotation but overlook the descriptions. It increases the difficulty in preparing large data sets when multiple datasets with divergent descriptions are involved. For example, the word for ‘hand; in Table 1 has various forms. I realised that /awəi/ means ‘hand’, ‘wrist’ and ‘arm’ after checking all the sources carefully. Another example is that English words like drink should mean  ‘the drink’ or the ‘to drink’ Although linguists should be aware of the ambiguity of these glosses, they still use these glosses without giving any further explanations.

4. Attribution:
   There have been several instances, where data that were made available in open access in the public domain have been used and/or reproduced without proper attribution of the source and without proper credits to the original collector of the data. This may be a deterrent to other linguists to make their data available openly.

In the light of the outlined problems, rendering the existing datasets comparable becomes a task that consumes a lot of time and energy, with a low chance of success, since the task cannot be done without thinking of novel data representations that still keep trace of the original sources.

Gloss	Reinhard 1970	Rana 2002	Watters 2005	Donohue  2013*	Aaley 2017
cloud	duliŋ		bəm ~ pãːyi / pãi	pãj  /pãj/	dʊliŋ
hand	tabi	nabi / amokh	awi / awəi	əwi /wi/	ɒmɒk, nabi, awi
fog	ganigiliŋ		dhundi		panji
to speak✢				məso /mso/	ɡipən ədʊ
to move (1st)✢	gaunʦən		ɡhu a-t-n̩, ɡho ə-ɡo		ɡhʊ əɡɒ

Table 1: The examples of existing data. ^* The format of Donohue’s data is in “broad” and /phonemic transcription/. ^✢  The glosses ‘to speak’ and ‘to move’ were not listed in the 200 items of the basic vocabulary list by Swadesh (1952).

New, comparable lexical data for Kusunda

Given the aforementioned problems, it was very nice to see the new data which Bodt And Aaley published  on Kusunda last year. They interviewed the (maybe) last two speakers Gyani Maiya and Kamala and published their lexical data freely online. The project was sponsored by the Endangered Language Fund (ELF), the CALC research grant, as well as a crowdfunding enterprise. The original recordings and a short paper describing the process and the data are published on Zenodo (10.5281/zenodo.3377537), where they can be freely downloaded.

Additionally,  a list of 250 basic vocabulary items (following the concept selection by Sagart et al. 2019) was prepared, archived with Zenodo, and presented in a short blog post (Aaley and Bodt 2020). In order to further enhance the comparability of the new Kusunda wordlist as published by Aaley and Bodt, we have now converted the original wordlist into CLDF format. The data in CLDF format themselves are curated on GitHub (https://github.com/lexibank/aaleykusunda, Version 1.0), and have also been archived with Zenodo (https://zenodo.org/record/3746946).

With the new data being available both in human- and machine-readable format, there is some hope that this fieldwork could inspire colleagues to start investigating the Kusunda language more closely or to help Aaley with his attempts to revitalize and document Kusunda. Additionally, the work addresses the four issues identified before:

1. Data storing and sharing:
   A short summary that describes the fieldwork method, along with the video and audio clips of the recordings are stored on Zenodo (https://zenodo.org/). Zenodo is a general-purpose open-access online archive, and it is widely used by scholars to deposit their dataset and articles. All data and articles submitted to this website are given Digital Object Identifiers (DOIs), and as long as the dataset remains online, the given DOI will always point to the corresponding datasets (and Zenodo has been made for long-time archiving, so we do not talk about the next five years here only). Therefore, the new Kusunda data are always traceable and will be very hard to erase from the internet.

2. Digitisation:
   Aaley and Bodt did not have time to analyse all data they shared, instead they shared them openly in the hope to trigger the interest of colleagues who could help in analysing the data further. The transcribed Kusunda vocabulary is stored in CLDF format (https://cldf.clld.org, Forkel et al. 2018) in a public GitHub repository (https://github.com/lexibank/aaleykusunda), so the lexical material can be accessed without restriction. Since GitHub repositories are always in flux, and data may change, and it is not guaranteed that a proprietary provider, such as GitHub, will guarantee long-term-archiving, distinct versions of the data are archived (again) with Zenodo. Data in CLDF format is provided in plain text form in form of comma-separated values (CSV) and can be viewed not only in text editors, but also with common spreadsheet software, such as Excel or Google Sheets. It can also be curated and analysed by several Python libraries (notably, LingPy, https://lingpy.org, List et al. 2019), and many of these libraries provide detailed instructions with many usage examples (see e.g. List et al. 2018).
3. Standardisation:
   The primary goal of CLDF is to standardise linguistic data for the purpose of cross-linguistic comparison. In order to give a better view of the CLDF format, figure 1 outlines a simplified CLDF structure along with the roles distributed between linguists and computer programs. Table 2 and Table 3 are brief examples drawn from the data.

1. 1. As shown in Table 2, each gloss is given a unique Concepticon gloss and a identification number (https://concepticon.clld.org, List et al. 2020). The clear definition can be found on the website. This strategy which regulates the usage of concepts can not only keep the data sheet “tidy”, but also gives a clear definition of the glosses.
   2. As shown in Table 3, the lexical items are transcribed in IPA. There are three columns to hold different versions of transcriptions. Usually, the first column (Value) preserves the lexical items in the raw data, the second column (Form) holds the sound sequences as they are given in slight automatically preprocessed form, and the last column displays the results after conversion and tokenization. In this way, the raw form, if not written in IPA, can be preserved and the last column gives the data for further analysis in software packages such as LingPy or annotation tools such as EDICTOR (https://digling.org/edictor/, List 2017).

4. Attribution:
   Like many academic articles often include a contribution section to detail the roles of each author, the repository of this dataset on Github also gives a CONTRIBUTORS.md file in which details the author (the field linguists), and the curator (also known as the maintainer). This keeps transparency of the source, and it shows appreciation to people who contribute their time and efforts in curating the data as well as keeping the data updated. Also, the new Kusunda dataset provides extensive instructions on how to cite the data in both bibtex format (the sources.csv in Figure 1) and the plain text form.

ID	Name	Concepticon_ID	Concepticon_Gloss	Definition (Concepticon)
25	the Cloud	1489	CLOUD	https://concepticon.clld.org/parameters/1489
74	to give	1447	GIVE	https://concepticon.clld.org/parameters/1447
81	the hand	1277	HAND	https://concepticon.clld.org/parameters/1277

Table 2: An example of the Concepticon dictionary (parameters.csv). The format includes parameters IDs (ID), the gloss (Name), the concepticon IDs (Concepticon_ID), and the concepticon concepts (Concepticon_Gloss). The links at the last column are not included in the data, as these links only provide convenience to retrieve the definition on the Concepticon website.

ID	Language_ID	Parameter_ID	Value	Form	Segments
KusundaGM-25-1	KusundaGM	25	bɐm	bɐm	b ɐ m
KusundaK-25-1	KusundaK	25	bɐm	bɐm	b ɐ m
Kusunda-25-1	Kusunda	25	bɐm	bɐm	b ɐ m
KusundaGM-74-1	KusundaGM	74	eː.gɔː	eː.gɔː	eː + g ɔː
KusundaK-74-1	KusundaK	74	eː.guː	eː.guː	eː + g uː
Kusunda-74-1	Kusunda	74	e	e	e
KusundaGM-81-1	KusundaGM	81	aː.wəj	aː.wəj	aː + w ə j
KusundaK-81-1	KusundaK	81	aː.wəj	aː.wəj	aː + w ə j
Kusunda-81-1	Kusunda	81	a.wəj	a.wəj	a + w ə j

Table 3: An example of the wordlist (forms.csv). The format includes unique entry IDs (ID), speaker or language IDs (Language_ID), the gloss ID (Parameter_ID, see the ID column in Table (2), the phonological rendering (Value), the phonetic sequence before tokenization (Form), and the tokenized phonetic sequences (Segments). The informants are Gyani Maiya (KusundaGM), Kamala (KusundaK), and the reconstructed proto-Kusunda words (Kusunda).

While people are complaining that linguistic data are not “open” enough, the way the new data are presented is a good example that shows that linguistic data can indeed be provided in transparent form. In addition the fieldwork done by Aaley and Bodt helps preserve the Kusunda culture, which is an important factor for a group’s ethnic identity. More work has to be done, and ideally, all recordings would be analysed and glossed, but it is obvious that this cannot be done immediately, but will require more time. Finally, the innovative ways used by Aaley and Bodt to obtain funding for their fieldwork via a crowdfunding campaign might also help to attract attention from a wider audience and encourage more scholars to work on language preservation.

I am very glad to see the new Kusunda data being presented openly on the internet, and I look forward to seeing more linguists to further work with the data, and come to new analyses as well as conclusions.

Acknowledgments 

I thank Timotheus A. Bodt，Yunfan Lai, Sandra Auderset and Johann-Mattis List for providing information and comments.

References

Aaley, Uday Raj. 2017. Kusunda Tribe and Dictionary. Uday Raj Aaley.

Uday Raj Aaley and Timotheus A. Bodt. 2020. “New Kusunda data: A list of 250 concepts,” in Computer-Assisted Language Comparison in Practice, 08/04/2020, https://calc.hypotheses.org/2414.

Forkel, Robert, Johann-Mattis List, Simon J. Greenhill, Christoph Rzymski,
Sebastian Bank, Michael Cysouw, Harald Hammarström, Martin Haspelmath,
Gereon A. Kaiping, and Russell D. Gray. 2018. “Cross-Linguistic Data Formats,
Advancing Data Sharing and Re-Use in Comparative Linguistics.” Scientific
Data 5 (1). https://doi.org/10.1038/sdata.2018.205.

Gautam, Bhojraj, Mark Donohue, and Madhav Pokharel. 2013. Kusunda Linguistics.
Canberra: Australian National University. http://kusunda.linguistics.anu.edu.au/wordlist/.

Hodgson, Brian Houghton. 1848a. “On the Chepang and Kusunda Tribes of
Nepal.” JASB 17 (1848): 650–58.

Johann-Mattis List,  Simon J. Greenhill, Tiago Tresoldi and Robert Forkel. 2019. “Lingpy/Lingpy:
A Python Library for Quantitative Tasks in Historical Linguistics. Version 2.6.5” Jena: Max Planck Institute for the Science of Human History. URL: https://lingpy.org

Johann-Mattis List, Christoph Rzymski, Simon J Greenhill, Nathanael E.
Schweikhard, Kristina Pianykh, Annika Tjuka, Mei-Shin Wu, and Robert Forkel.
2020. “Concepticon. Version 2.3.0.” Jena: Max Planck Institute for the Science of Human History. URL: https://concepticon.clld.org.

List, Johann-Mattis. 2017. “A Web-Based Interactive Tool for Creating, Inspecting,
Editing, and Publishing Etymological Datasets.” In Proceedings of the 15th
Conference of the European Chapter of the Association for Computational Linguistics.
System Demonstrations, 9–12. Valencia: Association for Computational
Linguistics. http://edictor.digling.org.

Johann-Mattis List, Mary Walworth, Simon J Greenhill, Tiago Tresoldi, and
Robert Forkel. 2018. “Sequence comparison in computational historical linguistics.”
Journal of Language Evolution 3 (2): 130–44. https://doi.org/10.1093/jole/lzy006.

Rana, B.K. 2002. “New Materials on the Kusunda Language.” Cambridge MA,
USA: Fourth Round Table International Conference on Ethnogenesis of South;
Central Asia; Harvard University.

Reinhard, Johan, and Sueyoshi Toba. 1970. A Preliminary Linguistic Analysis
and Vocabulary of the Kusunda Language. Kathmandu: Summer Institute of
Linguistics; Tribhuvan University.

Sagart, Laurent, Guillaume Jacques, Yunfan Lai, Robin Ryder, Valentin
Thouzeau, Simon J. Greenhill, and Johann-Mattis List. 2019. “Dated Language
Phylogenies Shed Light on the Ancestry of Sino-Tibetan.” Proceedings of the
National Academy of Science of the United States of America 116: 10317–22. DOI: 10.1073/pnas.1817972116.

Swadesh, Morris. 1952. “Lexico-Statistic Dating of Prehistoric Ethnic Contacts.
With Special Reference to North American Indians and Eskimos.” Proceedings
of the American Philosophical Society 96 (4): 452–63.

 

---

---