Under-resourced Languages Research Articles

Automatic authorship attribution in Albanian texts.

Automatic authorship identification is a challenging task that has been the focus of extensive research in natural language processing. Regardless of the progress made in attributing authorship, the need for corpora in under-resourced languages impedes advancing and examining present methods. To address this gap, we investigate the problem of authorship attribution in Albanian. We introduce a newly compiled corpus of Albanian newsroom columns and literary works and analyze machine-learning methods for detecting authorship. We create a set of hand-crafted features targeting various categories (lexical, morphological, and structural) relevant to Albanian and experiment with multiple classifiers using two different multiclass classification strategies. Furthermore, we compare our results to those obtained using deep learning models. Our investigation focuses on identifying the best combination of features and classification methods. The results reveal that lexical features are the most effective set of linguistic features, significantly improving the performance of various algorithms in the authorship attribution task. Among the machine learning algorithms evaluated, XGBoost demonstrated the best overall performance, achieving an F1 score of 0.982 on literary works and 0.905 on newsroom columns. Additionally, deep learning models such as fastText and BERT-multilingual showed promising results, highlighting their potential applicability in specific scenarios in Albanian writings. These findings contribute to the understanding of effective methods for authorship attribution in low-resource languages and provide a robust framework for future research in this area. The careful analysis of the different scenarios and the conclusions drawn from the results provide valuable insights into the potential and limitations of the methods and highlight the challenges in detecting authorship in Albanian. Promising results are reported, with implications for improving the methods used in Albanian authorship attribution. This study provides a valuable resource for future research and a reference for researchers in this domain.

A Case Study Applying a Framework of Contemporary Archival Theory to Lexicography Projects for Under-Resourced Languages

Archivy’s turn toward principles of inclusiveness and decolonization has been the topic of much recent scholarship (Cook, 2013; Lee, 2016). This ethical paradigm shift has been marked by an acknowledgment of the varied needs and desires of community stakeholders in archiving contexts, and ongoing efforts to explore the ways communities are participating in archiving their cultural heritage. There are parallels between this work and the preservation of endangered languages as one manifestation of cultural heritage although, traditionally, they are most readily observed in the domain of language archiving (Henke & Berez-Kroeker 2016; Wasson et al., 2016). This dissertation research initiates an exploration of community archiving principles that directly apply to lexicography (the creation of dictionaries). After conducting a rigorous scoping review of archival science literature and developing a framework of concepts from said literature that apply to lexicographic work, this research is pursuing a case study of four individual dictionaries for their respective languages (Frederick & Roeschley, 2024). Specifically, the case study aims to identify the extent to which the previously identified concepts of relationality and nontraditional conceptualizations apply to each dictionary project, the extent to which community archiving praxis is being utilized in each project, and the perceptions of each of the lexicographers regarding the applicability of the two aforementioned concepts to the scope of their work. An overarching goal of this research is to argue for enhanced collaboration and communication between two disciplines with similar histories and goals to create ethical and empowering outcomes for heritage communities.

A Deep Learning Automatic Speech Recognition Model for Shona Language

This study presented the development of a deep learning-based Automatic Speech Recognition (ASR) system for Shona, a low-resource language characterized by unique tonal and grammatical complexities. The research aimed to address the challenges posed by limited training data, a lack of labelled data, and the intricate tonal nuances present in Shona speech, with the objective of achieving significant improvements in recognition accuracy compared to traditional statistical models. Motivated by the limitations of existing approaches, the research addressed three key questions. Firstly, it explored the feasibility of using deep learning to develop an accurate ASR system for Shona. Secondly, it investigated the specific challenges involved in designing and implementing deep learning architectures for Shona speech recognition and proposed strategies to mitigate these challenges. Lastly, it compared the performance of the deep learning-based model with existing statistical models in terms of accuracy. The developed ASR system utilized a hybrid architecture consisting of a Convolutional Neural Network (CNN) for acoustic modelling and a Long Short-Term Memory (LSTM) network for language modelling. To overcome the scarcity of data, the research employed data augmentation techniques and transfer learning. Attention mechanisms were also incorporated to accommodate the tonal nature of Shona speech. The resulting ASR system achieved impressive results, with a Word Error Rate (WER) of 29%, Phoneme Error Rate (PER) of 12%, and an overall accuracy of 74%. These metrics indicated a significant improvement over existing statistical models, highlighting the potential of deep learning to enhance ASR accuracy for under-resourced languages like Shona. This research contributed to the advancement of ASR technology for under-resourced languages like Shona, ultimately fostering improved accessibility and communication for Shona speakers worldwide.

Analyzing the Effects of Transcription Errors on Summary Generation of Bengali Spoken Documents

Automatic speech recognition (ASR) has become an indispensable part of the AI domain, with various speech technologies reliant on it. The quality of speech recognition depends on the amount of annotated data used to train an ASR system, among other factors. For a low-resourced language, this is a severe constraint and thus ASR quality is often poor. Humans can read through text containing ASR-errors, provided the context of the sentence is preserved. Yet in cases of transcripts generated by ASR systems of low-resource languages, multiple important words are misrecognized and the context is mostly lost; discerning such a text becomes nearly impossible. This paper analyzes the types of transcription errors that occur while generating ASR transcripts of spoken documents in Bengali, an under-resourced language predominantly spoken in India and Bangladesh. The transcripts of the Bengali spoken document are generated using the ASR of Google Cloud Speech. The paper also explores if there is an effect of such transcription errors in generating speech summaries of these spoken documents. Summarization is carried out extractively; sentences are selected from the ASR-generated text of the spoken document. Speech summaries are created by aggregating the speech-segments from the original speech of the selected sentences. Subjective evaluation shows the ‘readability’ of the spoken summaries are not degraded by ASR errors, but the quality is affected due to the reliance on intermediate text-summary containing transcription errors.

The development of synthetic child speech in three South African languages

It is well-known that children with expressive communication difficulties have the right to communicate, but they should also have the right to do so in whichever language they choose, with a voice that closely matches their age, gender, and dialect. This study aimed to develop naturalistic synthetic child speech, matching the vocal identity of three children with expressive communication difficulties, using Tacotron 2, for three under-resourced South African languages, namely South African English (SAE), Afrikaans, and isiXhosa. Due to the scarcity of child speech corpora, 2 hours of child speech data per child was collected from three 11- to 12-year-old children. Two adult models were used to “warm start” the child speech synthesis. To determine the naturalness of the synthetic voices, 124 listeners participated in a mean opinion score survey (Likert Score) and optionally gave qualitative feedback. Despite limited training data used in this study, we successfully developed a synthesized child voice of adequate quality in each language. This study highlights that with recent technological advancements, it is possible to develop synthetic child speech that matches the vocal identity of a child with expressive communication difficulties in different under-resourced languages.

Scalability and diversity of StarGANv2-VC in Arabic emotional voice conversion: Overcoming data limitations and enhancing performance

Emotional Voice Conversion (EVC) for under-resourced languages like Arabic faces challenges due to limited emotional speech data. This study explored strategies to mitigate dataset scarcity and improve Arabic EVC performance. Fundamental experiments (Speaker-Dependent, Gender-Dependent, Gender-Independent) were conducted using the KSUEmotions dataset to analyze speaker, gender, and model impacts. Data augmentation techniques like time stretching and phase shuffling artificially increased data diversity. Attention mechanisms integrated into StarGANv2-VC aimed to better capture emotional cues. Transfer learning leveraged the larger English Emotional Speech Database (ESD) to enhance the Arabic system. A novel “Reordering Speaker-Emotion Data” approach treated each emotion as a separate speaker to expand the emotional variability. Our comprehensive approach, combining transfer learning, data augmentation, and architectural modifications, demonstrates the potential to overcome dataset limitations and enhance the performance of Arabic EVC systems.

Comparative Analysis of Deep Learning Models for Part of Speech Tagging in the Malay Language

Despite the widespread use of Malay, under-resourced languages like Malay face challenges in Natural Language Processing (NLP), particularly in Part-of-Speech (POS) tagging. The scarcity of annotated corpora poses a primary obstacle to POS tagging in Malay. This study aims to enhance the effectiveness and reliability of POS tagging models explicitly tailored for under-resourced languages within the field of NLP, focusing on Malay. Existing models, which rely on Conditional Random Fields and Hidden Markov Models, exhibit limitations, underscoring the need for more robust approaches. The research conducts a comparative analysis of various deep-learning models with different encoders for POS tagging in Malay sentences. The experimental analysis demonstrates that the Bidirectional Long Short-Term Memory (Bi-LSTM) model, leveraging a pre-trained Bidirectional Encoder Representations from Transformers (BERT) embedding model, achieves exceptional accuracy, precision, recall, and F1 scores in predicting tags. Notably, the BERT + Bi-LSTM model, boasting an accuracy of 98.82%, outperforms other models, showcasing superior performance across all evaluated metrics. Additionally, this combined model effectively handles known and unknown words, yielding highly accurate POS tagging results for Malay sentences. Doi: 10.28991/HIJ-2024-05-02-04 Full Text: PDF

Limitations of Large Language Models in Propaganda Detection Task

Propaganda in the digital era is often associated with online news. In this study, we focused on the use of large language models and their detection of propaganda techniques in the electronic press to investigate whether it is a noteworthy replacement for human annotators. We prepared prompts for generative pre-trained transformer models to find spans in news articles where propaganda techniques appear and name them. Our study was divided into three experiments on different datasets—two based on an annotated SemEval2020 Task 11 corpora and one on an unannotated subset of the Polish Online News Corpus, which we claim to be an even bigger challenge as an example of an under-resourced language. Reproduction of the results of the first experiment resulted in a higher recall of 64.53% than the original run, and the highest precision of 81.82% was achieved for gpt-4-1106-preview CoT. None of our attempts outperformed the baseline F1 score. One of the attempts with gpt-4-0125-preview on original SemEval2020 Task 11 achieved an almost 20% F1 score, but it was below the baseline, which oscillated around 50%. Part of our work that was dedicated to Polish articles showed that gpt-4-0125-preview had a 74% accuracy in the binary detection of propaganda techniques and 69% in propaganda technique classification. The results for SemEval2020 show that the outputs of generative models tend to be unpredictable and are hardly reproducible for propaganda detection. For the time being, these are unreliable methods for this task, but we believe they can help to generate more training data.

Data augmentation and transfer learning for cross-lingual Named Entity Recognition in the biomedical domain

AbstractGiven the increase in production of data for the biomedical field and the unstoppable growth of the internet, the need for Information Extraction (IE) techniques has skyrocketed. Named Entity Recognition (NER) is one of such IE tasks useful for professionals in different areas. There are several settings where biomedical NER is needed, for instance, extraction and analysis of biomedical literature, relation extraction, organisation of biomedical documents, and knowledge-base completion. However, the computational treatment of entities in the biomedical domain has faced a number of challenges including its high cost of annotation, ambiguity, and lack of biomedical NER datasets in languages other than English. These difficulties have hampered data development, affecting both the domain itself and its multilingual coverage. The purpose of this study is to overcome the scarcity of biomedical data for NER in Spanish, for which only two datasets exist, by developing a robust bilingual NER model. Inspired by back-translation, this paper leverages the progress in Neural Machine Translation (NMT) to create a synthetic version of the Colorado Richly Annotated Full-Text (CRAFT) dataset in Spanish. Additionally, a new CRAFT dataset is constructed by replacing 20% of the entities in the original dataset generating a new augmented dataset. We evaluate two training methods: concatenation of datasets and continuous training to assess the transfer learning capabilities of transformers using the newly obtained datasets. The best performing NER system in the development set achieved an F-1 score of 86.39%. The novel methodology proposed in this paper presents the first bilingual NER system and it has the potential to improve applications across under-resourced languages.

End-to-end automated speech recognition using a character based small scale transformer architecture

This study explores the feasibility of constructing a small-scale speech recognition system capable of competing with larger, modern automated speech recognition (ASR) systems in both performance and word error rate (WER). Our central hypothesis posits that a compact transformer-based ASR model can yield comparable results, specifically in terms of WER, to traditional ASR models while challenging contemporary ASR systems that boast significantly larger computational sizes. The aim is to extend ASR capabilities to under-resourced languages with limited corpora, catering to scenarios where practitioners face constraints in both data availability and computational resources. The model, comprising a compact convolutional neural network (CNN) and transformer architecture with 2.214 million parameters, challenges the conventional wisdom that large-scale transformer-based ASR systems are essential for achieving high accuracy. In comparison, contemporary ASR systems often deploy over 300 million parameters. Trained on a modest dataset of approximately 3000 h – significantly less than the 50,000 h used in larger systems – the proposed model leverages the Common Voice and LibriSpeech datasets. Evaluation on the LibriSpeech test-clean and test-other datasets produced character error rates (CERs) of 6.40% and 16.73% and WERs of 16.03% and 35.51% respectively. Comparisons with existing architectures showcase the efficiency of our model. A gated recurrent unit (GRU) architecture, albeit achieving lower error rates, incurred a computational cost 24 times larger than our proposed model. Large-scale transformer architectures, while achieving marginally lower WERs (2%–4% on LibriSpeech test-clean), require 200 times more parameters and 53,000 additional hours of training data. Modern large language models are used to improve the WERs, but require large computational resources. To further enhance performance, a small 4-g language model was integrated into our end-to-end ASR model, resulting in improved WERs. The overarching goal of this work is to provide a practical solution for practitioners dealing with limited datasets and computational resources, particularly in the context of under-resourced languages.

Ainu–Japanese Bi-directional Neural Machine Translation: A Step Towards Linguistic Preservation of Ainu, An Under-Resourced Indigenous Language in Japan

This study presents a groundbreaking approach to preserving the Ainu language, recognized as critically endangered by UNESCO, by developing a bi-directional neural machine translation (MT) system between Ainu and Japanese. Utilizing the Marian MT framework, known for its effectiveness with resource-scarce languages, the research aims to overcome the linguistic complexities inherent in Ainu's polysynthetic structure. The paper delineates a comprehensive methodology encompassing data collection from diverse Ainu text sources, meticulous preprocessing, and the deployment of neural MT models, culminating in the achievement of significant SacreBLEU scores that underscore the models' translation accuracy. The findings illustrate the potential of advanced MT technology to facilitate linguistic preservation and educational endeavors, advocating for integrating such technologies in safeguarding endangered languages. This research not only underscores the critical role of MT in bridging language divides but also sets a precedent for employing computational linguistics to preserve cultural and linguistic heritage.

Unearthing the latent assumptions inscribed into language tools: the cross-cultural benefits of applying a reflexive lens in co-design

ABSTRACT Language technology tools provide a promising way to teach, share, retain, and curate under-resourced language learning materials in community. The inclusion of language teachers working with communities increases the potential for designed tools to be adopted by those groups. However, there is little research concerning the adaptation of tools designed with one community to other languages. To identify the implications for such scalability, we ran workshops with the ‘Record and Write’ tool, developed as a versatile format for collection, curation, and use of under-resourced language learning materials in community. The process enabled teachers of languages with varying availability of teaching materials to reflect on some of the embedded complexities of adapting the tool to their context. This paper critically reflects on the design process of the tool and design lessons learned relating to language governance, the reflection of culture in database tools, conversational learning support, and differentiated needs for grammatical accuracy and annotation. Methodologically, the paper proposes a reflexive lens on co-design in cross-cultural contexts, identifying some of the latent assumptions embedded in technologies that emerge when tools are transposed to different language and learning contexts.

Bridging the Gap

Effective organization and retrieval of news content are heavily reliant on accurate news classification. While the mountainous research has been conducted in resourceful languages like English and Chinese, the researches on under-resourced languages like the Kurdish language are severely lacking. To address this challenge, we introduce a hybrid approach called RFO-CNN in this paper. The proposed method combines an improved version of red fox optimization algorithm (RFO) and convolutional neural network (CNN) for finetuning CNN’s parameters. Our model’s efficacy was tested on two widely used Kurdish news datasets, KNDH and KDC-4007, both of which contain news articles classified into various categories. We compared the performance of RFO-CNN to other cutting-edge deep learning models such as bidirectional long short-term memory networks and bidirectional encoder representations from transformers (BERT) transformers, as well as classical machine learning approaches such as multinomial naive bayes, support vector machine, and K-nearest neighbors. We trained and tested our datasets using four different scenarios: 60:40, 70:30, 80:20, and 90:10. Our experimental results demonstrate the superiority of the RFO-CNN model across all scenarios, outperforming the benchmark BERT model and other machine learning models in terms of accuracy and F1-score.

Switching Self-Attention Text Classification Model with Innovative Reverse Positional Encoding for Right-to-Left Languages: A Focus on Arabic Dialects

Transformer models have emerged as frontrunners in the field of natural language processing, primarily due to their adept use of self-attention mechanisms to grasp the semantic linkages between words in sequences. Despite their strengths, these models often face challenges in single-task learning scenarios, particularly when it comes to delivering top-notch performance and crafting strong latent feature representations. This challenge is more pronounced in the context of smaller datasets and is particularly acute for under-resourced languages such as Arabic. In light of these challenges, this study introduces a novel methodology for text classification of Arabic texts. This method harnesses the newly developed Reverse Positional Encoding (RPE) technique. It adopts an inductive-transfer learning (ITL) framework combined with a switching self-attention shared encoder, thereby increasing the model’s adaptability and improving its sentence representation accuracy. The integration of Mixture of Experts (MoE) and RPE techniques empowers the model to process longer sequences more effectively. This enhancement is notably beneficial for Arabic text classification, adeptly supporting both the intricate five-point and the simpler ternary classification tasks. The empirical evidence points to its outstanding performance, achieving accuracy rates of 87.20% for the HARD dataset, 72.17% for the BRAD dataset, and 86.89% for the LABR dataset, as evidenced by the assessments conducted on these datasets.

Language Varieties of Italy: Technology Challenges and Opportunities

Abstract Italy is characterized by a one-of-a-kind linguistic diversity landscape in Europe, which implicitly encodes local knowledge, cultural traditions, artistic expressions, and history of its speakers. However, most local languages and dialects in Italy are at risk of disappearing within a few generations. The NLP community has recently begun to engage with endangered languages, including those of Italy. Yet, most efforts assume that these varieties are under-resourced language monoliths with an established written form and homogeneous functions and needs, and thus highly interchangeable with each other and with high-resource, standardized languages. In this paper, we introduce the linguistic context of Italy and challenge the default machine-centric assumptions of NLP for Italy’s language varieties. We advocate for a shift in the paradigm from machine-centric to speaker-centric NLP, and provide recommendations and opportunities for work that prioritizes languages and their speakers over technological advances. To facilitate the process, we finally propose building a local community towards responsible, participatory efforts aimed at supporting vitality of languages and dialects of Italy.

AC-IQuAD: Automatically Constructed Indonesian Question Answering Dataset by Leveraging Wikidata

AbstractConstructing a question-answering dataset can be prohibitively expensive, making it difficult for researchers to make one for an under-resourced language, such as Indonesian. We create a novel Indonesian Question Answering dataset that is produced automatically end-to-end. The process uses Context Free Grammar, the Wikipedia Indonesian Corpus, and the concept of the proxy model. The dataset consists of 134 thousand simple questions and 60 thousand complex questions. It achieved competitive grammatical and model accuracy compared to the translated dataset but suffers from some issues due to resource constraints.

Ethio-Fake: Cutting-Edge Approaches to Combat Fake News in Under-Resourced Languages Using Explainable AI

The proliferation of fake news has emerged as a significant threat to the integrity of information dissemination, particularly on social media platforms. Misinformation can spread quickly due to the ease of creating and disseminating content, affecting public opinion and sociopolitical events. Identifying false information is therefore essential to reducing its negative consequences and maintaining the reliability of online news sources. Traditional approaches to fake news detection often rely solely on content-based features, overlooking the crucial role of social context in shaping the perception and propagation of news articles. In this paper, we propose a comprehensive approach that integrates social context-based features with news content features to enhance the accuracy of fake news detection in under-resourced languages. We perform several experiments utilizing a variety of methodologies, including traditional machine learning, neural networks, ensemble learning, and transfer learning. Assessment of the outcomes of the experiments shows that the ensemble learning approach has the highest accuracy, achieving a 0.99 F1 score. Additionally, when compared with monolingual models, the fine-tuned model with the target language outperformed others, achieving a 0.94 F1 score. We analyze the functioning of the models, considering the important features that contribute to model performance, using explainable AI techniques.

Improving the Performance of Low-resourced Speaker Identification with Data Preprocessing

Automatic speaker identification is done to tackle daily security problems. Speech data collection is an essential but very challenging task for under-resourced languages like Burmese. The speech quality is crucial to accurately recognize the speaker’s identity. This work attempted to find the optimal speech quality appropriate for Burmese tone to enhance identification compared with other more richy resourced languages on Mel-frequency cepstral coefficients (MFCCs). A Burmese speech dataset was created as part of our work because no appropriate dataset available for use. In order to achieve better performance, we preprocessed the foremost recording quality proper for not only Burmese tone but also for nine other Asian languages to achieve multilingual speaker identification. The performance of the preprocessed data was evaluated by comparing with the original data, using a time delay neural network (TDNN) together with a subsampling technique that can reduce time complexity in model training. The experiments were investigated and analyzed on speech datasets of ten Asian languages to reveal the effectiveness of the data preprocessing. The dataset outperformed the original dataset with improvements in terms of equal error rate (EER). The evaluation pointed out that the performance of the system with the preprocessed dataset improved that of the original dataset.

Homophobia and transphobia detection for low-resourced languages in social media comments

People are increasingly sharing and expressing their emotions using online social media platforms such as Twitter, Facebook, and YouTube. An abusive, hateful, threatening, and discriminatory act that makes discomfort targets gay, lesbian, transgender, or bisexual individuals is called Homophobia and Transphobia. Detecting these types of acts on social media is called Homophobia and Transphobia Detection. This task has recently gained interest among researchers. Identifying homophobic and transphobic content for under-resourced languages is a bit challenging task. There are no such resources for Malayalam and Hindi to categorize these types of content as far now. This paper presents a new high-quality dataset for detecting homophobia and transphobia in Malayalam and Hindi languages. Our dataset consists of 5,193 comments in Malayalam and 3,203 comments in Hindi. We also submitted the experiments performed with traditional machine learning and transformer-based deep learning models on the Malayalam, Hindi, English, Tamil, and Tamil-English datasets.

Shahmukhi named entity recognition by using contextualized word embeddings

Named Entity Recognition (NER) is an imperative Natural Language Processing (NLP) task which intents to identify and classify predefined named entities in a given span of text. For many Western and Asian languages, NER is a systematically premeditated and established task, however, a little work has been done for Shahmukhi. This paper presents Shahmukhi NER with four key contributions. First, a Bi-directional Long-Short Term Memory (BiLSTM) network based NER model has been developed by incorporating various features including character and word embeddings and Part of Speech (POS) tagging. Second, transfer learning has been employed by training context-free Word2Vec and contextualized Embeddings from Language Models (ELMo) word representations. The word representations have been trained using a Shahmukhi corpus of 14.9 million words. Third, we prepared a cleaner version of an existing Shahmukhi NER corpus by performing Unicode normalization and tokenization errands. The corpus has been deduplicated and results are reported on an unseen evaluation set which produced valid results. Fourth, we have studied the impact of two annotation schemes; Inside-Outside (IO) and Inside-Outside-Beginning (IOB) for Shahmukhi. Transfer learning was quite helpful to enhance the performance of NER models especially ELMo embeddings significantly improved the results by prompting contextualized embedding vectors. This is the first study to use character embeddings, POS tagging and transfer learning for Shahmukhi named entity recognition. The IO scheme based model achieved an accuracy of 98.60% with an f-score of 83.75. The IOB scheme based model performed with an accuracy of 98.43% and an f-score of 75.55. These scores are quite promising for an under-resourced morphologically-rich language.