The Livonians are a Finno-Ugric nation indigenous to Latvia. They are presently the most endangered culture in the European Union and their language is one of the most endangered languages in the world. The Livonian speech community lost its last compactly inhabited territory during the Soviet occupation of Latvia and since then has experienced a steady but continuous loss of Livonian language speakers. Today, only approximately 20 people are able to hold a conversation in Livonian. While presently living scattered across Latvia, with their language no longer used as a medium of daily interaction, the Livonian community has taken this challenging reality and turned it to its advantage in its approach to creating language resources from a limited set of human and archival resources. This article is devoted specifically to the subject of digital humanities in the service of endangered languages and cultures. Understanding the specifics of endangered and especially critically endangered languages and cultures, which actually form the majority of the world's cultural and linguistic diversity, is critically important both in building tailor-made digital resources for such communities and applying ready-made instruments for their benefit. This article offers a broader picture of the challenges and opportunities that developers of digital resources for critically endangered languages and cultures have to take into account, based on the Livonian experience. The article summarises and presents some of the findings and ideas deriving from the creation of digital resources and tools for the Livonian language and culture, approaches for overcoming a lack of resources and limited data issues, as well benefits and experiences that low resource - especially extremely low resource - cultures and languages can derive from the Livonian experience.
The Livonians are a Finno-Ugric nation indigenous to Latvia. Currently, approximately
20 individuals are able to communicate in Livonian [
The Livonian community has been able to preserve its identity and also its language
up to the present day despite its small size (19th century – 2400 [
Though the Livonian language is currently considered to be one of the world’s most
endangered languages and is listed in the UNESCO Atlas of the World’s Languages in
Danger as critically endangered [
This article is devoted specifically to the subject of digital humanities in the service of endangered languages and cultures. Understanding the specifics of such languages and cultures, which actually form the majority of the world’s cultural and linguistic diversity, is critically important both in building tailor-made digital resources for such communities and applying ready-made instruments for their benefit. Our article is intended to offer a broader picture of the challenges and opportunities that developers of resources for critically endangered languages and cultures have to take into account, based on the Livonian experience. For this purpose we wish to summarise and present some of the findings and ideas, approaches for overcoming a lack of resources and limited data issues, as well benefits and experiences that low resource – especially extremely low resource – cultures and languages can derive from the Livonian experience. 2
When considering digital technologies as an opportunity for all languages to survive
(especially endangered ones), it is important to remember that for most languages in
1 Although it incorrectly states that the last first-language speaker died in 2009.
the world, inclusion in the digital domain remains a distant possibility. However, the
good news is that the same technologies that empower more widespread languages can
also empower the less widespread – or even endangered – ones. Although for
endangered languages, this work should be carried out with special effort and advocacy [
The general challenges to be taken into account when building digital resources for low-resource languages and cultures are widely known – insufficient institutional, financial, political support for research and development, a lack of human resources, and especially a lack of skilled personnel. Another challenge is limited data leading to fewer automation options when building digital resources and tools. However, when it comes to languages and cultures, which are extremely small in size, there are additional issues to be taken into account.
First of all, any resources compiled or collected in an endangered language with very
few speakers cannot be effectively processed or used without overcoming the language
barrier. While the involvement of skilled personnel is already limited for low-resource
cultures, finding such personnel with sufficient proficiency in an endangered language
limits any effort even further. One of the ways of overcoming this issue – as is done in
the Livonian case – is to involve the speaker community, utilising them not only as
language consultants, but also involving them in the creation, processing and revision
of content. Such an approach of course helps to create valuable linguistic resources;
however, supervision of this process by linguists and specialists to ensure quality of the
material is a necessity [
This involvement means that resources have to be useful and beneficial not only for the researchers, but also for the language and culture community itself, which is an important target group of potential users for any such resources. It is important, however, that while in most cases resources for endangered cultures are created with an intent to help to ensure the sustainability, preservation, and development or even revitalisation of such cultures, such work is ineffective if carried out without close cooperation with the community and responding to its crucial needs.
For the Livonian language, this is evidenced by a long history of compiling Livonian
dictionaries. The first major Livonian dictionaries were already published in the
mid19th century [
Trying to match the outputs of the resources to community needs can also present additional challenges. An example of this is a UL Livonian Institute project dedicated to language acquisition. During the course of this project, a survey was conducted to understand the Livonian-language acquisition needs in the community with an eye towards including these needs in tools based on resources developed for the Livonian language and culture.
Nicholas Ostler [
In studying the language acquisition needs (as well as other aspects of language
acquisition) of the community, some general issues emerge relating to the theoretical
facets of language acquisition, pedagogy, language standardisation as well as practical
issues affecting the processes of language maintenance, development, and possibilities
for use. In-depth interviews with good Livonian speakers (B and C language
proficiency level) and questionnaires completed by other Livonian community members
showed a particular interest and emphasis on appropriate digital solutions for language
use or ‘language on the computer’ as one area where there was particular interest and
hope for expanding the use of Livonian in real life and fostering language transfer to
the younger generation. It is important to note that respondents used the concept
‘language on the computer’ to mean different digital language resources (from dictionaries
and the possibility to check for a correct word form, to the expanded availability of
language corpora, texts, information about and in Livonian, etc.) and less to language
in other media, which is important for younger people, but must be further researched
from the perspective of endangered languages and the effect on language practice [
The complex nature of this process can be seen more clearly when analysing just one aspect of Livonian language revitalisation, namely creating language learning and teaching materials (for Livonian as a second language or L2, or as a heritage language). This task seems basic and straightforward from the theoretical perspective of larger languages used in daily life, but it becomes more complex when such materials have to be created for a language like Livonian, where challenges are not only related to linguistic characteristics but maybe even more to sociolinguistic issues.
Creating such materials reveals a variety of challenges: an absence of already existing pedagogical materials or an accompanying tradition of such materials to build on, a lack of pedagogically-trained personnel as well as insufficient language proficiency among potential teachers in general, a lack of language acquisition programmes for teachers and the wider community, an absence of certain domains of language use accompanied by an instant need for new terminology to fill these gaps, a lack of standards, which are central to creating learning materials, improving accessibility, and more.
Thus, the smaller a language or culture community, the more aspects actually have to be taken into account when developing even the simplest resources.
Although the first attempts to create digital resources for Livonian already took place
in the 1990s (e.g. Livonian-Latvian-Livonian dictionary [
Creation of existing resources was initially entirely data and community driven. It
started with the need to have an online Livonian dictionary which could be easily
updated and expanded, serving not only as a collection of vocabulary, but also as a
reference for the Livonian literary standard (currently the open-access dictionary of
standardised Livonian contains ca. 12500 lemmas). Due to the fact that Livonian has
extremely rich morphology, dictionary lemmas were then supplemented with a database
of morphological forms (all forms and their variations of declinable words). The need
to acquire additional vocabulary from texts, including textual parts of the
LivonianEstonian-Latvian Dictionary [
Though still under development, these resources have already significantly simplified the process for learning about and studying the Livonian language and culture, and have created a foundation for future solutions directed towards simplifying access to Livonian-related materials. Existing resources have been serving as a base for contemporary research and subsequently also the standardisation of Livonian, and currently an overview of Livonian grammar is being prepared based on data acquired from these databases. In 2020, work started on further development of the cluster by adding three important datasets and expanding it with cultural data. As the corpus is not only a source for linguistic data, but also a collection of texts (folklore, culture texts, newspaper texts), the cluster is supplemented with background data for the sources it uses, with an option for the inclusion of data describing future sources such as ethnographic collections, etc. Similarly, background data on people (including language informants and owners of original items) is added. Both of these are also necessary from the perspective of achieving better quality and results from the research.
Limited data does not permit the use of statistics as a deciding factor in many cases, when research is performed in specific areas. In languages and cultures with sufficient data, statistics normally point towards the most common and regular use, but for endangered cultures, which have limited data and are regularly exposed to majority cultures, this may not always be the case. If, for example, two forms of a word are different out of a total of ten recorded, it might well be that these two are actually correct and the rest of the cases are mistakes caused by the specific backgrounds of each informant or for other reasons. Therefore, data on the circumstances and authors of each source becomes more important, the smaller the community being researched. This, however, raises the question of how to ensure data anonymity when researching endangered cultures.
The third database currently under development – a geospatially linked place name database – is formally an extension of the lexical database. The creation of this database currently includes ca. 191 Livonian toponyms (over 1000 expected by the summer of 2021) and was initially also community driven due to practical considerations relating to collecting Livonian place names, so that bilingual road signs could be introduced in the territories historically inhabited by the Livonians.
Creating this resource has been quite challenging. There is no single Livonian place name archive and place names are primarily found in lexicographic publications or studies, albeit in a more scattered manner and fewer in number [12, 16, 20, etc.]. Currently, recording a large number of place names and localising them through field work is also not possible, as there remain only very few Livonian speakers and – due to the historical situation of the Livonians – their connection with the territories inhabited by their ancestors is often indirect and their knowledge of Livonian place names is meagre.
Thus, in large part, the Livonian language corpus is also used to acquire data for this resource as well as metadata from various collections that points to the place of collection, including folklore collections, ethnographic collections, etc. located in different institutions in Latvia and abroad. Mapping the place names is very important, as these places not only form a layer of linguistic data, but also reflect locations of specific importance to Livonian culture. These places are often not found in any available cartographic product (for example, homesteads which have disappeared or have been moved, drained rivers or lakes, etc.). Mapping itself is performed using several approaches, including textual study, comparison with corresponding Latvian place names that are in many cases borrowed or translated from Livonian, etc.
Fig. 5. Screenshot from the place name module (tool for adding the geospatial information). This resource is particularly important in expanding possibilities for research and access to other Livonian language and culture resources. The mapped place name database is planned to serve as a starting point for the creation of a one-stop database cluster and its backbone. This will make it possible to link place names, geospatial data, and information from other fields (e.g., information on informants or original possessors of sources, dialect materials, objects, folklore, oral history collections, etc.), thereby opening up new possibilities for the multifaceted documentation and research of Livonian heritage in the future. Furthermore, this will make it possible to use data for many different purposes beyond research. These could include, for example, digital exhibits or using the cartographic products resulting from this project for education exploring regional heritage, cultural tourism, development of municipal strategies, entrepreneurship, and many other areas. The need for such outputs can already be seen in various numerous requests by state and municipal institutions, private entrepreneurs and NGOs that the UL Livonian Institute is trying to answer. 3.3
One of the benefits that endangered languages and cultures can have from possessing only limited data is data integration. While larger communities tend to develop their resources across several institutions, applying different principles and thus struggling with the need to link data among them, endangered cultures are capable of handling data in a more consistent way. This is a key data principle integrated into Livonian resources that also ensures better data extraction and handling later on, especially in light of insufficient human resources.
In resources developed for Livonian, each kind of data is kept separate, but is used in other datasets as linked data. Thus, the lexical database only contains data on lemmas and meanings, but no textual units – these are located in the corpus. Even example material found in the Livonian-Estonian-Latvian Dictionary – forms separate text in the corpus; examples in the current dictionary are featured as corpus links. The same applies to the place name database, which contains geospatial data, but the place names it features as lemmas are linked from the lexical database along with examples from the corpus and grammatical information from the morphological database. This approach ensures that adjustments or alterations made in one dataset are reflected correctly in other resources without any additional effort.
Another important – and inevitable – principle of Livonian-language resources is multilingualism. In order for Livonian intangible heritage (linguistic data, folklore, cultural texts, etc.) to be available to researchers, the Livonian community, and also the general public, which in large part has no proficiency in Livonian, translations have to be provided for almost every textual unit. The primary languages selected for translations are Latvian – as the main language used by the Livonian community and the general public in Latvia – and English for international users. However, as the Livonian language and culture has been documented and researched in a number of countries, there is also a point in using already existing translations in, e.g., Estonian, Finnish, German, etc.
The Livonian experience also shows that when developing digital resources for such
tiny communities, manual work is inevitable and only some processes can be fully
entrusted to automated solutions, at least in their initial phases. For example, automated
text recognition would not be effective since most of the texts are handwritten or printed
at a poor level of quality. Also, as only a small part of them are available electronically,
automated indexing does not work, because of a lack of clear and verified grammar
rules and limited data, etc. [
As there are also few sufficiently proficient to work with intangible heritage data in an endangered language, the question of how to acquire resources and increase the efficiency of existing ones becomes crucial. In the Livonian case, this issue is addressed by both technical and political solutions. On the technical side, this is done first by lessening the amount of work through the use of data handling and data extraction principles, e.g., when corpus data is indexed, it is also added to the morphological and lexical databases.
Resource efficiency is also solved by workflow matching different skill levels – those less skilled can perform simple tasks that tend to form the majority of data processing, while complex tasks are left for those more skilled. This comes along with a principle of providing access to data as soon as it is added to a database. This also ensures that dynamic adjustments can be made by those processing or using data for their research. Because of this approach, it is also possible to use crowd sourcing done by researchers, where those performing research using Livonian resources may contribute at every stage of their work by eliminating spotted mistakes, adding missing indexation, translation, and so on during the course of their research.
To help solve the problem of an insufficient number of personnel proficient in Livonian available to work on the resources, a political solution has also been suggested. The UL Livonian Institute is currently implementing a pilot project and has offered the idea to launch a programme for language speaker community involvement. In this programme, language speakers would be paid for working on various tasks involving language, including processing (indexing, transcribing texts) or adding data (e.g., voice data for the dictionary) to the digital resources under development. 4
The emergence of digital technologies provides novel possibilities for preserving and providing access to endangered language data. Up to now this data has been too inaccessible for linguists and remaining language speakers alike, and it is important that digital resources are created in close cooperation between both of these groups. As a result, there is mutual benefit by helping to ensure better sustainability for the endangered communities, while also significantly increasing the social impact of research results.
Although endangered cultures suffer greatly from a lack of resources, many issues can be resolved using strategic and carefully considered approaches to the structure of resources and workflows. In addition, sometimes even political solutions can be as important as technical features. At the same time, having limited data may be beneficial for endangered cultures, as this may permit the creation of more integrated and consistent resources, thereby compensating for the lack of quantity with increased quality of data processing.
Though Livonian resources were initially created as a Livonian language data archive and a tool for language research, standardisation and acquisition, its principles have been expanded to create the cultural resources currently under development. In the future, these resources can also be further adjusted to suit other types of digital data such as images, audio recordings, videos, 3D scans, data from other databases, thereby offering a whole new perspective for studying and preserving one of the most endangered cultures in Europe.
The products and discoveries resulting from this resource may also be useful to other smaller communities, and the synergy and coordination among various resources can create a rich, high-quality resource suitable for multifaceted studies in many fields, or for interdisciplinary research in general.
This study was supported by the Latvian Council of Science Fundamental and Applied Research Programme project “Documenting and mapping Livonian place names and creating an official place name register” (LZP-2019/1-0240), the National Research Programme project “Digital Resources for Humanities: Integration and Development” (VPP-IZM-DH-2020/1-0001), and the PostDoc Latvia project “Applying the Common European Framework of Reference for Languages to Livonian: A New Opportunity for Endangered Languages” (1.1.1.2/VIAA/3/19/527).