TaTa: A Multilingual Table-to-Text Dataset for African Languages

Abstract

Existing data-to-text generation datasets are mostly limited to English. To address this lack of data, we create Table-to-Text in African languages (TATA), the first large multilingual table-to-text dataset with a focus on African languages. We created TATA by transcribing figures and accompanying text in bilingual reports by the Demographic and Health Surveys Program, followed by professional translation to make the dataset fully parallel. TATA includes 8,700 examples in nine languages including four African languages (Hausa, Igbo, Swahili, and Yorùbá) and a zero-shot test language (Russian). We additionally release screenshots of the original figures for future research on multilingual multi-modal approaches. Through an in-depth human evaluation, we show that TATA is challenging for current models and that less than half the outputs from an mT5-XXL-based model are understandable and attributable to the source data. We further demonstrate that existing metrics perform poorly for TATA and introduce learned metrics that achieve a high correlation with human judgments. 1

Cross-lingual Open-Retrieval Question Answering for African Languages

“…In parallel with efforts to include more low-resource languages in NLP research (Costajussà et al, 2022;Ruder, 2020), demand for NLP that targets African languages, which represent more than 30% of the world's spoken languages (Ogueji et al, 2021) is growing. This has resulted in the creation of publicly available multilingual datasets targeting African languages for a variety of NLP tasks such as sentiment analysis (Muhammad et al, 2023;Shode et al, 2022), language identification (Adebara et al, 2022), datato-text generation (Gehrmann et al, 2022), topic classification (Adelani et al, 2023;Hedderich et al, 2020), machine translation (Adelani et al, 2022a;Nekoto et al, 2020), and NER (Eiselen, 2016;Adelani et al, 2021Adelani et al, , 2022b.…”