TaTA (Table-to-Text in African languages) is the first large multilingual table-to-text datasets with a focus on African languages. The dataset is parallel and covers nine languages, eight of which are spoken in Africa: Arabic, English, French, Hausa, Igbo, Portuguese, Swahili, Yorùbá, and Russian.
TaTa was created by extracting tables from charts in 71 PDF reports published between 1990 and 2021 by the Demographic and Health Surveys Program. The reports are published in English and commonly a second language. The extracted tables were then translated from English by professional translators to all eight languages. The final dataset comprises 8,479 tables. For more information on the dataset creation, refer to the paper.