Researchers at the Council for Scientific and Industrial Research (CSIR) are in the research phase of developing an app that helps foundation-phase children to develop literacy skills in their home language, and to learn to read, speak and understand the language.

While still in the research phase, the CSIR Natural Language Processing Research Group had applied for additional funding and hoped to get the app onto the Play Store as soon as possible to enable people to download and use it, said CSIR Natural Language Processing Research Group senior researcher Dr Laurette Marais.

The app, run on entry-level tablet devices during the series of four, ten-week pilot studies, presents sentences of the appropriate grammatic complexity for their level to the learners.

It displays a sentence, or uses text-to-speech processing capabilities to pronounce a sentence, and children select from tokens that represent different phonetic groupings based on teaching and learning techniques specific to the language to reconstruct the sentence.

The child could then also pronounce the sentence, with the app's natural language processing capabilities grading this against the correct pronunciation, but also based on dialect differences, she explained.

Further, the learner can also be asked to translate a sentence in his or her home language into another language, which was English during the pilot phase, or from another language into his or her home language.

This helps to grade the child's comprehension of the words and the sentence, CSIR researchers said on January 28.

“Home language literacy is important, as research has shown that it empowers and liberates people, helps to reduce poverty by enabling inclusion and promotes equality, because it is an equaliser,” said Marais.

“Knowledge is power and a home language is one of the best vehicles for imparting knowledge and enabling people to learn and develop themselves. Children need to understand what they are reading or hearing, and this is important to address if we want to promote equality,” she said.

The context in which children learned was important to understand to shape an effective solution. The indigenous African languages spoken in South Africa differed from one another, including how words were written and what constituted a word, among others, she explained.

Therefore, techniques used to teach English, for example, do not necessarily translate well to teaching South Africa's other languages.

The teaching technique needed to be done in such a way that it fitted the language technique, and solutions needed to be developed that directly addressed the literacy needs of children in their home languages, she emphasised.

The research group worked closely with teachers, including what the impact of the learners' interactions with the app was and what was important to teachers, she added.

The app currently only had four of South Africa's 11 spoken languages, namely isiZulu, Sepedi, English and Afrikaans. The researchers aim to broaden this to include all 11 official spoken languages, said Marais.

Additionally, the group is looking at, in some cases, implementing some of the capabilities of the app to run only on a device, without the need for an Internet connection.

“We are working with other [CSIR] research groups that have expertise in network development in rural areas to roll out the solutions and, as part of our two-pronged approach, we are working to transfer some of the capabilities to devices in further iterations,” she said.

Further, the research group had developed 11 text-to-speech systems for all 11 official spoken languages, and these solutions enabled the bridging of communication barriers to allow citizens to access and communicate in their preferred languages, when previously they were not normally able to do so, said CSIR Natural Language Processing Research Group acting research group leader and principal specialist Avi Moodley.

The group is also in the research and development phase of enabling each of the text-to-speech voices to speak multiple languages, which is important in South Africa's multilingual culture where people may speak multiple languages, sometimes within the same sentence.

“This capability to speak multiple languages would not only be beneficial to government departments to cater to different language needs, but would also be beneficial for private and corporate companies to represent their singular brand across multiple languages when engaging with their customer bases,” he said.

“We are actively seeking partners, including private-sector partners, to sponsor the costs associated with rolling out the technology, such as having a telecommunications company sponsor some of the devices learners need for these initiatives, for example,” he said.

The CSIR also has a portfolio of related natural language products and offerings that government or private companies can deploy, and it can provide its expertise in designing, developing and implementing large language models.

The CSIR could also provide playbooks for companies looking to implement such models in their businesses, which were based on the learnings the research group had gained over nearly two decades of working with natural language models, said Moodley.

The CSIR research group had also identified the potential use of the language app in higher education, Marais added.

This project is funded by the Department of Sport, Arts and Culture, and is called Ngiyaqonda!, which is isiZulu for “I understand”.