In this paper we present the speech corpus for the Siberian Ingrian Finnish language. The speech corpus includes audio data, annotations, software tools for data-processing, two databases and a web application. We have published part of the audio data and annotations. The software tool for parsing annotation files and feeding a relational database is developed and published under a free license. A web application is developed and available. At this moment, about 300 words and 200 phrases can be displayed using this web application.

Focus on language-specific properties with insights from formal minimalist syntax can improve universal dependency (UD) parsing. Such improvements are especially sensitive for low-resource African languages, like Wolof, which have fewer UD treebanks in number and amount of annotations, and fewer contributing annotators. For two different UD parser pipelines, one parser model was trained on the original Wolof treebank, and one was trained on an edited treebank. For each parser pipeline, the accuracy of the edited treebank was higher than the original for both the dependency relations and dependency labels. Accuracy for universal dependency relations improved as much as 2.90%, while accuracy for universal dependency labels increased as much as 3.38%. An annotation scheme that better fits a language’s distinct syntax results in better parsing accuracy.

Language technologies, particularly speech technologies, are becoming more pervasive for access to digital platforms and resources. This brings to the forefront concerns of their inclusivity, first in terms of language diversity. Additionally, research shows speech recognition to be more accurate for men than for women and more accurate for individuals younger than 30 years of age than those older. In the Global South where languages are low resource, these same issues should be taken into consideration in data collection efforts to not replicate these mistakes. It is also important to note that in varying contexts within the Global South, this work presents additional nuance and potential for bias based on accents, related dialects and variants of a language. This paper documents i) the designing and execution of a Linguists Engagement for purposes of building an inclusive Kiswahili Speech Recognition dataset, representative of the diversity among speakers of the language ii) the unexpected yet key learning in terms of socio-linguistcs which demonstrate the importance of multi-disciplinarity in teams developing datasets and NLP technologies iii) the creation of a test dataset intended to be used for evaluating the performance of Speech Recognition models on demographic groups that are likely to be underrepresented.

Language revitalisation should not be understood as a direct outcome of language documentation, which is mainly focused on the creation of language repositories. Natural language processing (NLP) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages. In this paper, we discuss the current state of the interaction between language documentation and computational linguistics, present a diagnosis of how the outputs of recent documentation projects for endangered languages are underutilised for the NLP community, and discuss how the situation could change from both the documentary linguistics and NLP perspectives. All this is introduced as a bridging paradigm dubbed as Computational Language Documentation and Development (CLD²). CLD² calls for (1) the inclusion of NLP-friendly annotated data as a deliverable of future language documentation projects; and (2) the exploitation of language documentation databases by the NLP community to promote the computerization of endangered languages, as one way to contribute to their revitalization.

Data augmentation strategies are increasingly important in NLP pipelines for low-resourced and endangered languages, and in neural morphological inflection, augmentation by so called data hallucination is a popular technique. This paper presents a detailed analysis of inflection models trained with and without data hallucination for the low-resourced Canadian Indigenous language Gitksan. Our analysis reveals evidence for a concatenative inductive bias in augmented models—in contrast to models trained without hallucination, they strongly prefer affixing inflection patterns over suppletive ones. We find that preference for affixation in general improves inflection performance in “wug test” like settings, where the model is asked to inflect lexemes missing from the training set. However, data hallucination dramatically reduces prediction accuracy for reduplicative forms due to a misanalysis of reduplication as affixation. While the overall impact of data hallucination for unseen lexemes remains positive, our findings call for greater qualitative analysis and more varied evaluation conditions in testing automatic inflection systems. Our results indicate that further innovations in data augmentation for computational morphology are desirable.

Many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs. One bottleneck is the time-intensive nature of annotation. An even narrower bottleneck occurs for recordings with access constraints, such as language that must be vetted or filtered by authorised community members before annotation can begin. We propose a privacy-preserving workflow to widen both bottlenecks for recordings where speech in the endangered language is intermixed with a more widely-used language such as English for meta-linguistic commentary and questions (e.g.What is the word for ‘tree’?). We integrate voice activity detection (VAD), spoken language identification (SLI), and automatic speech recognition (ASR) to transcribe the metalinguistic content, which an authorised person can quickly scan to triage recordings that can be annotated by people with lower levels of access. We report work-in-progress processing 136 hours archival audio containing a mix of English and Muruwari. Our collaborative work with the Muruwari custodian of the archival materials show that this workflow reduces metalanguage transcription time by 20% even given only minimal amounts of annotated training data, 10 utterances per language for SLI and for ASR at most 39 minutes, and possibly as little as 39 seconds.

This paper describes the motivation and implementation details for a rule-based, index-preserving grapheme-to-phoneme engine ‘G_i2P_i' implemented in pure Python and released under the open source MIT license. The engine and interface have been designed to prioritize the developer experience of potential contributors without requiring a high level of programming knowledge. ‘G_i2P_i' already provides mappings for 30 (mostly Indigenous) languages, and the package is accompanied by a web-based interactive development environment, a RESTful API, and extensive documentation to encourage the addition of more mappings in the future. We also present three downstream applications of ‘G_i2P_i' and show results of a preliminary evaluation.

Accelerating the process of data collection, annotation, and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages (Bird, 2009). Our experiments describe how we maximize the quality for the Nepal Bhasa syntactic complement structure chunking model. Native speaker language consultants were trained to annotate a minimally selected raw data set (Suárez et al.,2019). The embedded clauses, matrix verbs, and embedded verbs are annotated. We apply both statistical training algorithms and transfer learning in our training, including Naive Bayes, MaxEnt, and fine-tuning the pre-trained mBERT model (Devlin et al., 2018). We show that with limited annotated data, the model is already sufficient for the task. The modeling resources we used are largely available for many other endangered languages. The practice is easy to duplicate for training a shallow parser for other endangered languages in general.

We describe recent extensions to the open source Learning And Reading Assistant (LARA) supporting image-based and phonetically annotated texts. We motivate the utility of these extensions both in general and specifically in relation to endangered and archaic languages, and illustrate with examples from the revived Australian language Barngarla, Icelandic Sign Language, Irish Gaelic, Old Norse manuscripts and Egyptian hieroglyphics.

In this paper we present an approach to efficiently recover texts from corrupted documents of endangered languages. Textual resources for such languages are scarce, and sometimes the few available resources are corrupted PDF documents. Endangered languages are not supported by standard tools and present even the additional difficulties of not possessing any corpus over which to train language models to assist with the recovery. The approach presented is able to fully recover born digital PDF documents with minimal effort, thereby helping the preservation effort of endangered languages, by extending the range of documents usable for corpus building.

Transcribing speech for primarily oral, local languages is often a joint effort involving speakers and outsiders. It is commonly motivated by externally-defined scientific goals, alongside local motivations such as language acquisition and access to heritage materials. We explore the task of ‘learning through transcription’ through the design of a system for collaborative speech annotation. We have developed a prototype to support local and remote learner-speaker interactions in remote Aboriginal communities in northern Australia. We show that situated systems design for inclusive non-expert practice is a promising new direction for working with speakers of local languages.

This paper discusses the development of a Part-of-Speech tagger for te reo Māori which is the Indigenous language of Aotearoa, also known as New Zealand, see Morrison. Henceforth, Part-of-Speech will be referred to as POS throughout this paper and te reo Māori will be referred to as Māori, while Universal Dependencies will be referred to as UD. Prior to the development of this tagger, there was no POS tagger for Māori from Aotearoa. POS taggers tag words according to their syntactic or grammatical category. However, many traditional syntactic categories, and by consequence POS labels, do not “work for” Māori. By this we mean that, for some of the traditional categories, The definition of, or guidelines for, an existing category is not suitable for Māori. They do not have an existing category for certain word classes of Māori. They do not reflect a Māori worldview of the Māori language. We wanted a tagset that is usable with industry-wide tools, but we also needed a tagset that would meet the needs of Māori. Therefore, we based our tagset and guidelines on the UD tagset and tagging conventions, however the categorization of words has been significantly altered to be appropriate for Māori. This is because at the time of development of our POS tagger, the UD conventions had still not been used to tag a Polyneisan language such as Māori, nor did it provide any guidelines about how to tag them. To that end, we worked with highly-proficient, specially-selected Māori speakers and linguists who are specialists in Māori. This has ensured that our POS labels and guidelines conventions faithfully reflect a Māori speaker’s conceptualization of their language.

Innu-Aimun is an Algonquian language spoken in Eastern Canada. It is the language of the Innu, an Indigenous people that now lives for the most part in a dozen communities across Quebec and Labrador. Although it is alive, Innu-Aimun sees important preservation and revitalization challenges and issues. The state of its technology is still nascent, with very few existing applications. This paper proposes a first survey of the available linguistic resources and existing technology for Innu-Aimun. Considering the existing linguistic and textual resources, we argue that developing language technology is feasible and propose first steps towards NLP applications like machine translation. The goal of developing such technologies is first and foremost to help efforts in improving language transmission and cultural safety and preservation for Innu-Aimun speakers, as those are considered urgent and vital issues. Finally, we discuss the importance of close collaboration and consultation with the Innu community in order to ensure that language technologies are developed respectfully and in accordance with that goal.

This paper describes how emerging linguistic resources and technologies can be used to build a language learning platform for Irish, an endangered language. This platform, An Scéalaí, harvests learner corpora - a vital resource both to study the stages of learners’ language acquisition and to guide future platform development. A technical description of the platform is provided, including details of how different speech technologies and linguistic resources are fused to provide a holistic learner experience. The active continuous participation of the community, and platform evaluations by learners and teachers, are discussed.

For decades, researchers in natural language processing and computational linguistics have been developing models and algorithms that aim to serve the needs of language documentation projects. However, these models have seen little use in language documentation despite their great potential for making documentary linguistic artefacts better and easier to produce. In this work, we argue that a major reason for this NLP gap is the lack of a strong foundation of application software which can on the one hand serve the complex needs of language documentation and on the other hand provide effortless integration with NLP models. We further present and describe a work-in-progress system we have developed to serve this need, Glam.

Machine translation for low-resource languages, such as Guarani, is a challenging task due to the lack of data. One way of tackling it is using pretrained word embeddings for model initialization. In this work we try to check if currently available data is enough to train rich embeddings for enhancing MT for Guarani and Spanish, by building a set of word embedding collections and training MT systems using them. We found that the trained vectors are strong enough to slightly improve the performance of some of the translation models and also to speed up the training convergence.

In this paper, we present a game with a purpose (GWAP) (Von Ahn 2006). The aim of the game is to promote language learning and ‘noticing’ (Skehan, 2013). The game has been designed for Irish, but the framework could be used for other languages. Irish is a minority language which means that L2 learners have limited opportunities for exposure to the language, and additionally, there are also limited (digital) learning resources available. This research incorporates game development, language pedagogy and ICALL language materials development. This paper will focus on the language materials development as this is a bottleneck in the teaching and learning of minority and endangered languages.

Many endangered Uralic languages have multilingual machine readable dictionaries saved in an XML format. However, the dictionaries cover translations very inconsistently between language pairs, for instance, the Livonian dictionary has some translations to Finnish, Latvian and Estonian, and the Komi-Zyrian dictionary has some translations to Finnish, English and Russian. We utilize graph-based approaches to augment such dictionaries by predicting new translations to existing and new languages based on different dictionaries for endangered languages and Wiktionaries. Our study focuses on the lexical resources for Komi-Zyrian (kpv), Erzya (myv) and Livonian (liv). We evaluate our approach by human judges fluent in the three endangered languages in question. Based on the evaluation, the method predicted good or acceptable translations 77% of the time. Furthermore, we train a neural prediction model to predict the quality of the automatically predicted translations with an 81% accuracy. The resulting extensions to the dictionaries are made available on the online dictionary platform used by the speakers of these languages.

Grammar checkers (GEC) are needed for digital language survival. Very low resource languages like Lule Sámi with less than 3,000 speakers need to hurry to build these tools, but do not have the big corpus data that are required for the construction of machine learning tools. We present a rule-based tool and a workflow where the work done for a related language can speed up the process. We use an existing grammar to infer rules for the new language, and we do not need a large gold corpus of annotated grammar errors, but a smaller corpus of regression tests is built while developing the tool. We present a test case for Lule Sámi reusing resources from North Sámi, show how we achieve a categorisation of the most frequent errors, and present a preliminary evaluation of the system. We hope this serves as an inspiration for small languages that need advanced tools in a limited amount of time, but do not have big data.

There are many challenges in morphological fieldwork annotation, it heavily relies on segmentation and feature labeling (which have both practical and theoretical drawbacks), it’s time-intensive, and the annotator needs to be linguistically trained and may still annotate things inconsistently. We propose a workflow that relies on unsupervised and active learning grounded in Word-and-Paradigm morphology (WP). Machine learning has the potential to greatly accelerate the annotation process and allow a human annotator to focus on problematic cases, while the WP approach makes for an annotation system that is word-based and relational, removing the need to make decisions about feature labeling and segmentation early in the process and allowing speakers of the language of interest to participate more actively, since linguistic training is not necessary. We present a proof-of-concept for the first step of the workflow, in a realistic fieldwork setting, annotators can process hundreds of forms per hour.

This is a report on results obtained in the development of speech recognition tools intended to support linguistic documentation efforts. The test case is an extensive fieldwork corpus of Japhug, an endangered language of the Trans-Himalayan (Sino-Tibetan) family. The goal is to reduce the transcription workload of field linguists. The method used is a deep learning approach based on the language-specific tuning of a generic pre-trained representation model, XLS-R, using a Transformer architecture. We note difficulties in implementation, in terms of learning stability. But this approach brings significant improvements nonetheless. The quality of phonemic transcription is improved over earlier experiments; and most significantly, the new approach allows for reaching the stage of automatic word recognition. Subjective evaluation of the tool by the author of the training data confirms the usefulness of this approach.

The project XXXX is developing a platform to enable researchers of living languages to easily create and make available state-of-the-art spoken and textual annotated resources. As a case study we use Greek and Pomak, the latter being an endangered oral Slavic language of the Balkans (including Thrace/Greece). The linguistic documentation of Pomak is an ongoing work by an interdisciplinary team in close cooperation with the Pomak community of Greece. We describe our experience in the development of a Latin-based orthography and morphologically annotated text corpora of Pomak with state-of-the-art NLP technology. These resources will be made openly available on the XXXX site and the gold annotated corpora of Pomak will be made available on the Universal Dependencies treebank repository.

This study investigates applications of automatic speech recognition (ASR) techniques to Hupa, a critically endangered Native American language from the Dene (Athabaskan) language family. Using around 9h12m of spoken data produced by one elder who is a first-language Hupa speaker, we experimented with different evaluation schemes and training settings. On average a fully connected deep neural network reached a word error rate of 35.26%. Our overall results illustrate the utility of ASR for making Hupa language documentation more accessible and usable. In addition, we found that when training acoustic models, using recordings with transcripts that were not carefully verified did not necessarily have a negative effect on model performance. This shows promise for speech corpora of indigenous languages that commonly include transcriptions produced by second-language speakers or linguists who have advanced knowledge in the language of interest.