A South African-built language technology tool is helping researchers develop artificial-intelligence systems for languages that do not have the large datasets available to English and other widely used languages.
IOL reports that TextAugment, created by Professor Vukosi Marivate and Tshephisho Sefara, uses data-augmentation techniques to expand limited language datasets for natural-language processing. The software has recorded more than 286,000 downloads and has been used in work involving languages including Swahili, Arabic and Uzbek.
The project won the inaugural NSTF-SADiLaR Research Software Award in July. Its developers say tools of this kind can make AI systems more representative by giving researchers practical ways to work with languages that have historically been overlooked by major technology platforms.
Source: IOL
