Hasil untuk "African languages and literature"

Lukas Radosky, Ivan Polasek

Generative artificial intelligence attracts significant attention, especially with the introduction of large language models. Its capabilities are being exploited to solve various software engineering tasks. Thanks to their ability to understand natural language and generate natural language responses, large language models are great for processing various software documentation artifacts. At the same time, large language models excel at understanding structured languages, having the potential for working with software programs and models. We conduct a literature review on the usage of large language models for software engineering tasks related to documentation and modeling. We analyze articles from four major venues in the area, organize them per tasks they solve, and provide an overview of used prompt techniques, metrics, approaches to human-based evaluation, and major datasets.

Gabrial Zencha Ashungafac, Mardhiyah Sanni, Busayo Awobade et al.

Recent advances in speech-enabled AI, including Google's NotebookLM and OpenAI's speech-to-speech API, are driving widespread interest in voice interfaces globally. Despite this momentum, there exists no publicly available application-specific model evaluation that caters to Africa's linguistic diversity. We present AfriSpeech-MultiBench, the first domain-specific evaluation suite for over 100 African English accents across 10+ countries and seven application domains: Finance, Legal, Medical, General dialogue, Call Center, Named Entities and Hallucination Robustness. We benchmark a diverse range of open, closed, unimodal ASR and multimodal LLM-based speech recognition systems using both spontaneous and non-spontaneous speech conversation drawn from various open African accented English speech datasets. Our empirical analysis reveals systematic variation: open-source ASR models excels in spontaneous speech contexts but degrades on noisy, non-native dialogue; multimodal LLMs are more accent-robust yet struggle with domain-specific named entities; proprietary models deliver high accuracy on clean speech but vary significantly by country and domain. Models fine-tuned on African English achieve competitive accuracy with lower latency, a practical advantage for deployment, hallucinations still remain a big problem for most SOTA models. By releasing this comprehensive benchmark, we empower practitioners and researchers to select voice technologies suited to African use-cases, fostering inclusive voice applications for underserved communities.

Tsosheletso Chidi, Nompumelelo Zondi, Gabi Mkhize

Black queer feminist literature remains under-researched. This reflects the societal marginalisation of black queer authors in South Africa. Our article offers a comparative analysis of the representation of black queer women by black queer and cisgender authors in selected isiXhosa and English poetry. The poems selected are from Unam Wena (2021) by Mthunzikazi Mbungwana and red cotton (2018) by vangile gantsho. Firstly, we explore how queer feminism is captured from a Xhosa perspective. Secondly, we explore how English is used to expose readers to black queerness, and, thirdly, we question how literary scholarship influences or limits black queer feminist literature and the functionality of queer feminist poetry as representations of black women. Discourse theory is used to examine how authors of the selected poetry construct knowledge about black queerness from a feminist perspective and shape how people understand it. In this article we adopt a narrative enquiry within the constructionism paradigm with qualitative textual analysis. Our analysis of the poetry reveals that, although the selected poets use two different languages, the same protest voice is foregrounded, with observable differences being primarily technical—namely how form, sound, and structure are employed to set the tone and mood in the issues addressed.

Patricia Alonso-Alvarez, Nees Jan van Eck

Unlike traditional proprietary data sources such as Scopus and the Web of Science (WoS), OpenAlex emphasizes its comprehensiveness. This study analyzes OpenAlex coverage and metadata completeness and accuracy of African research publications. To achieve this, OpenAlex is compared with Scopus, WoS, and African Journals Online (AJOL). First, we examine the coverage of African research publications in OpenAlex relative to Scopus, WoS, and AJOL. Then, we assess and compare the availability and accuracy of metadata in OpenAlex, Scopus, and WoS. The findings indicate that OpenAlex offers the most extensive publication coverage. In terms of metadata, OpenAlex provides high coverage for publication and author information, though its coverage of affiliations, references, and funder information is comparatively lower. Metadata accuracy is similarly high for publication and author fields, while affiliation, reference, and funding information show higher rates of missing or incomplete data. Notably, the results demonstrate that both metadata availability and accuracy in OpenAlex improve significantly for publications also indexed in Scopus and WoS. These findings suggest that OpenAlex has the potential to replace proprietary data sources for certain types of analyses. However, for some metadata fields, there remains a trade-off between extensiveness and accuracy.

Anastasia Zhukova, Christian E. Matt, Bela Gipp

Domain-specific languages that use a lot of specific terminology often fall into the category of low-resource languages. Collecting test datasets in a narrow domain is time-consuming and requires skilled human resources with domain knowledge and training for the annotation task. This study addresses the challenge of automated collecting test datasets to evaluate semantic search in low-resource domain-specific German language of the process industry. Our approach proposes an end-to-end annotation pipeline for automated query generation to the score reassessment of query-document pairs. To overcome the lack of text encoders trained in the German chemistry domain, we explore a principle of an ensemble of "weak" text encoders trained on common knowledge datasets. We combine individual relevance scores from diverse models to retrieve document candidates and relevance scores generated by an LLM, aiming to achieve consensus on query-document alignment. Evaluation results demonstrate that the ensemble method significantly improves alignment with human-assigned relevance scores, outperforming individual models in both inter-coder agreement and accuracy metrics. These findings suggest that ensemble learning can effectively adapt semantic search systems for specialized, low-resource languages, offering a practical solution to resource limitations in domain-specific contexts.

Mofetoluwa Adeyemi, Akintunde Oladipo, Ronak Pradeep et al.

Large language models (LLMs) have shown impressive zero-shot capabilities in various document reranking tasks. Despite their successful implementations, there is still a gap in existing literature on their effectiveness in low-resource languages. To address this gap, we investigate how LLMs function as rerankers in cross-lingual information retrieval (CLIR) systems for African languages. Our implementation covers English and four African languages (Hausa, Somali, Swahili, and Yoruba) and we examine cross-lingual reranking with queries in English and passages in the African languages. Additionally, we analyze and compare the effectiveness of monolingual reranking using both query and document translations. We also evaluate the effectiveness of LLMs when leveraging their own generated translations. To get a grasp of the effectiveness of multiple LLMs, our study focuses on the proprietary models RankGPT-4 and RankGPT-3.5, along with the open-source model, RankZephyr. While reranking remains most effective in English, our results reveal that cross-lingual reranking may be competitive with reranking in African languages depending on the multilingual capability of the LLM.

E. Gnugesser, C. Chwila, S. Brenner et al.

Background and Objectives Hypertension is one of the leading cardiovascular risk factors with high numbers of undiagnosed and untreated patients in Sub Saharan Africa (SSA). The health systems and affected people are often overwhelmed by the social and economic burden that comes with the disease. However, the research on the economic burden and consequences of hypertension treatment remains scare in SSA. The objective of our review was to compare different hypertension treatment costs across the continent and identify major cost drivers. Material and Methods Systematic literature searches were conducted in multiple databases (e.g., PubMed, Web of Science, Google Scholar) for peer reviewed articles written in English language with a publication date from inception to Jan. 2022. We included studies assessing direct and indirect costs of hypertension therapy in SSA from a provider or user perspective. The search and a quality assessment were independently executed by two researchers. All results were converted to 2021 US Dollar. Results Of 3999 results identified in the initial search, 33 were selected for data extraction. Costs differed between countries, costing perspectives and cost categories. Only 25% of the SSA countries were mentioned in the studies, with Nigeria dominating the research with a share of 27% of the studies. We identified 15 results each from a user or provider perspective. Medication costs were accountable for the most part of the expenditures with a range from 1.70$ to 97.06$ from a patient perspective and 0.09$ to 193.55$ from a provider perspective per patient per month. Major cost drivers were multidrug treatment, inpatient or hospital care and having a comorbidity like diabetes. Conclusion Hypertension poses a significant economic burden for patients and governments in SSA. Interpreting and comparing the results from different countries and studies is difficult as there are different financing methods and cost items are defined in different ways. However, our results identify medication costs as one of the biggest cost contributors. When fighting the economic burden in SSA, reducing medication costs in form of subsidies or special interventions needs to be considered. Trial registration Registration: PROSPERO, ID CRD42020220957.

H. Viviers, Rikus Ruben De Villiers, N. van der Merwe

ABSTRACT This study measures the levels of self-efficacy beliefs to determine how this correlates with academic success in introductory tertiary accounting within a South African context. Also, self-efficacy beliefs are compared to determine if significant differences exist based on gender, academic language, type of study funding and different professional programmes studied. The study applies social cognitive theory and gathers quantitative data using questionnaires with statistical analysis to determine self-efficacy belief levels in a first-year accounting module. The means of several variables were compared via t-tests or ANOVAs. Hereafter, regression analysis was applied to determine if self-efficacy beliefs significantly impact academic performance. Findings indicate that most first-year accounting students felt confident in their self-efficacy beliefs. Significant differences were observed in students’ self-efficacy beliefs based on gender, academic language and type of programme studied, but not based on type of study funding. Several self-efficacy beliefs correlated significantly with academic performance, though differences were observed for different types of programmes. As limited consideration of the impact of self-efficacy beliefs within the discipline of accounting prevails in the literature, and even less so in South Africa, the study serves to enhance accounting educators’ understanding on how self-efficacy beliefs impact academic performance at the first-year level.

Maria Mushaathoni, Madoda Cekiso

Motivation and students’ attitudes are generally cited as two of the most important factors influencing language-learning performance. Motivation and attitude not only enhance students’ understanding of learning and make them more open to learning, but also raise their expectations of the learning process and lower their anxiety. Therefore, the current study sought to establish the university students’ perceptions of their lecturers’ motivational strategies and their attitudes towards the English language, and the English Academic module. The study was quantitative in nature and a survey research design was followed. A questionnaire was used to solicit data from 150 first-year Foundation students who were conveniently selected. Likert items were used to measure the respondents’ perceptions of their lecturers’ motivational strategies and their attitudes toward the English Language and English Academic module. To analyse data, the most frequent responses are considered by working out percentages that agree, disagree, etc. The findings showed that the students perceived their lecturers as using a range of motivational strategies. These include, but are not limited to, allowing students to ask questions, providing feedback, motivating students to read more material, and praising students for good learning behaviour. The findings further revealed that the majority of students had a positive attitude towards the English language and English Academic module. Lecturers’ motivational strategies and students’ attitudes towards the English language and English Academic module could be perceived as significant considerations for lecturers when they tailor instruction and module design to the needs of the students. In addition, lecturers’ awareness of students’ attitudes can serve as vital information that could be used as a springboard to change the students’ attitudes for the better.

Francine Maessen, Bibi Burger, Mathilda Smit

In this review article, we focus on the depiction of the transnational and translingual as a state of being in-between in Klara du Plessis’s debut poetry collection, Ekke (2018). This in-between state has implications for how identity, place and visual art feature in the collection. Ekke contains fragments of German and French, but consists mainly of English interspersed with Afrikaans. The creation of meaning through this linguistic slippage reflects the idea of identity as always in-process that comes to the fore throughout the collection. Ekke also represents an intervention in South African urban literature, as Bloemfontein, a city not much featured in literature, is represented in several poems. In these poems, the poet/speaker struggles to situate Bloemfontein and its surrounding areas’ histories and symbolism in the transnational networks that she is a part of. The conception of identity and language being constantly in-progress is also conveyed in the collection’s poems about visual art. In these poems, meaning is created through the interaction of language with visual art, a process the poet calls ‘intervisuality'.

Benjamin Biggs, Ian McInerney, Eric C. Kerrigan et al.

The growing proliferation of FPGAs and High-level Synthesis (HLS) tools has led to a large interest in designing hardware accelerators for complex operations and algorithms. However, existing HLS toolflows typically require a significant amount of user knowledge or training to be effective in both industrial and research applications. In this paper, we propose using the Julia language as the basis for an HLS tool. The Julia HLS tool aims to decrease the barrier to entry for hardware acceleration by taking advantage of the readability of the Julia language and by allowing the use of the existing large library of standard mathematical functions written in Julia. We present a prototype Julia HLS tool, written in Julia, that transforms Julia code to VHDL. We highlight how features of Julia and its compiler simplified the creation of this tool, and we discuss potential directions for future work.

David Ifeoluwa Adelani, Jesujoba Oluwadara Alabi, Angela Fan et al.

Recent advances in the pre-training of language models leverage large-scale datasets to create multilingual models. However, low-resource languages are mostly left out in these datasets. This is primarily because many widely spoken languages are not well represented on the web and therefore excluded from the large-scale crawls used to create datasets. Furthermore, downstream users of these models are restricted to the selection of languages originally chosen for pre-training. This work investigates how to optimally leverage existing pre-trained models to create low-resource translation systems for 16 African languages. We focus on two questions: 1) How can pre-trained models be used for languages not included in the initial pre-training? and 2) How can the resulting translation models effectively transfer to new domains? To answer these questions, we create a new African news corpus covering 16 languages, of which eight languages are not part of any existing evaluation dataset. We demonstrate that the most effective strategy for transferring both to additional languages and to additional domains is to fine-tune large pre-trained models on small quantities of high-quality translation data.

R. Zakari, Zaharaddeen Karami Lawal, Idris Abdulmumin

The processing of natural languages is an area of computer science that has gained growing attention recently. NLP helps computers recognize, in other words, the ways in which people use their language. NLP research, however, has been performed predominantly on languages with abundant quantities of annotated data, such as English, French, German and Arabic. While the Hausa Language is Africa's second most commonly used language, only a few studies have so far focused on Hausa Natural Language Processing (HNLP). In this research paper, using a keyword index and article title search, we present a systematic analysis of the current literature applicable to HNLP in the Google Scholar database from 2015 to June 2020. A very few research papers on HNLP research, especially in areas such as part-of-speech tagging (POS), Name Entity Recognition (NER), Words Embedding, Speech Recognition and Machine Translation, have just recently been released. This is due to the fact that for training intelligent models, NLP depends on a huge amount of human-annotated data. HNLP is now attracting researchers' attention after extensive research on NLP in English and other languages has been performed. The key objectives of this paper are to promote research, to define likely areas for future studies in the HNLP, and to assist in the creation of further examinations by researchers for relevant studies.

A. Yallew, P. Langa, N. Nkhoma

Abstract One of the features of the internationalisation of higher education is the increasing use of the English language as a medium of instruction and research in higher education. This growing use of English spurs this article’s attempt at systematically reviewing literature focusing on African higher education systems. The analysis of the selected literature focuses on assessing the main themes, theoretical assumptions and core findings. As a result, 30 articles accessed through continental and international research databases were included in the final analysis after a five-step selection process using relevant keywords related to the topic and the context of the study. The findings indicate that research on the English language in higher education in African contexts overwhelmingly focuses on the language as a medium of teaching and learning. The other aspects, such as the role that the language plays as a medium of research and archiving knowledge, seem to be overlooked. The findings of the majority of both empirical and review papers seem to present critical and, at times, unfavourable views on the role English plays in the specific contexts studied. In light of these findings, the recommendation is that the role of the English language as a medium of instruction should be expanded to cover issues related to research, publication and archiving knowledge. This indicates that the continent’s higher education systems need more research on English language, which suggests that robust and pragmatic theoretical approaches might also be needed in future studies. A further observation is that the findings from the reviewed studies might be the result of using theories that are underpinned in traditions that are already critical of the use of the English language. Thus, more research attention could be given to strengthening the efficacy of using multiple theoretical perspectives to render the African contexts studied more intelligible.

Mavis Adu Asuamah, Raymond Boateng

A growing body of research has explored the healthcare experiences of African migrants in China. However, within this extant literature, there is a lacuna on the reproductive healthcare experiences of African women within this population. This study adopts semi-structured in-depth interviews in exploring the challenges to reproductive healthcare access  among African women in Beijing. Results indicate that African women face multiple barriers to accessing reproductive healthcare. In particular, the absence of reproductive health awareness, discriminatory immigration policy, discontentment with healthcare services, and language barrier were the key challenges identified. The study highlights the challenges of reproductive healthcare experiences among African migrant women in Beijing, China, and recommends the implementation of secure and equitable policies that cater for the needs of African women and minorities in the healthcare setting.

Thuli M. Makhura, Gary W. Collins, Hendrietta Segabutla et al.

Background: South Africa has incorporated Computer-Assisted Language Learning (CALL) into many university classrooms in order to help address ubiquitous concerns related to the limited English language proficiency of first-year university students. In the context of this study, the research site used the CALL application called the English Word Power (EWP programme). Research to establish students’ perceptions of the CALL application is somewhat limited, although students’ perceptions of a learning environment can be more useful in explaining their behaviour. Therefore, teachers’ understanding of the students’ perceptions towards a new learning programme is likely to assist them in tailoring the content according to the needs of the learners. Objectives: The objective of this study was to explore the students’ perceptions of the use of the EWP, a computer-based programme used for improving English language proficiency. The researchers endeavoured to gain an understanding of the students’ perceptions of the strengths and frustrations of the EWP programme. Specifically, the focus was on what the students perceived as their preferred mode of learning and what their views were regarding the contribution of the EWP programme in improving their language skills. Method: The study was qualitative in nature and a case study design was adopted. A purposive sample of 60 students from an Information and Communication Technology’s (ICT) Foundation programme was used to collect data. Focus group interviews were conducted with 60 purposefully selected students and content analysis was used to process data. Results: The study yielded mixed results, as some students were happy with some of the components of the EWP programme whilst others were frustrated with some components. Specifically, some students reported experiencing frustration with the irrelevant nature of the EWP content. On the contrary, the study revealed that students were satisfied with the EWP programme’s accessibility, which facilitated their learning opportunities. Some students reported that their language proficiency concerning spelling and vocabulary had improved. The findings further revealed that the students preferred face-to-face learning to the EWP programme learning environment. Conclusion: The implication of the findings is that students need a language learning programme that is relevant to their field of study. They also require a programme that allows for student–student and lecturer–student interaction.

Harm de Wet, Vukosi Marivate

Disinformation through fake news is an ongoing problem in our society and has become easily spread through social media. The most cost and time effective way to filter these large amounts of data is to use a combination of human and technical interventions to identify it. From a technical perspective, Natural Language Processing (NLP) is widely used in detecting fake news. Social media companies use NLP techniques to identify the fake news and warn their users, but fake news may still slip through undetected. It is especially a problem in more localised contexts (outside the United States of America). How do we adjust fake news detection systems to work better for local contexts such as in South Africa. In this work we investigate fake news detection on South African websites. We curate a dataset of South African fake news and then train detection models. We contrast this with using widely available fake news datasets (from mostly USA website). We also explore making the datasets more diverse by combining them and observe the differences in behaviour in writing between nations' fake news using interpretable machine learning.

Caleb Helbling, Fırat Aksoy

The difficulty associated with storing closures in a stack-based environment is known as the funarg problem. The funarg problem was first identified with the development of Lisp in the 1970s and hasn't received much attention since then. The modern solution taken by most languages is to allocate closures on the heap, or to apply static analysis to determine when closures can be stack allocated. This is not a problem for most computing systems as there is an abundance of memory. However, embedded systems often have limited memory resources where heap allocation may cause memory fragmentation. We present a simple extension to the prenex fragment of System F that allows closures to be stack-allocated. We demonstrate a concrete implementation of this system in the Juniper functional reactive programming language, which is designed to run on extremely resource limited Arduino devices. We also discuss other solutions present in other programming languages that solve the funarg problem but haven't been formally discussed in the literature.

D. Nsagha, C. W. Ayima, Theophile Nana-Njamen et al.

S. Vurayai, A. Ndofirepi

ABSTRACT This theoretical paper uses content analysis by drawing from the literature on the implications of the varied conceptions of ‘Publish or perish’ maxim to novice African academics in the neoliberal context. Neoliberal politics has defined education and research in particular as a commodity to be bought and sold as controlled by the profit motive and related market forces. As a result, novice African academics have been marginalised in the production and dissemination of knowledge on a global scale due to weapons such as cultural imperialism, linguicide, epistemicide and poor funding. We recommend that African academics should be independent of the private economic and political interests and serve the public interest in promoting their language, culture, and knowledge through their journals.