Multilingual pre-trained language models (PLMs) have demonstrated impressive performance on several downstream tasks on both high resourced and low-resourced languages. However, there is still a large performance drop for languages unseen during pre-training, especially African languages. One of the most effective approaches to adapt to a new language is language adaptive fine-tuning (LAFT) -- fine-tuning a multilingual PLM on monolingual texts of a language using the same pre-training objective. However, African languages with large monolingual texts are few, and adapting to each of them individually takes large disk space and limits the cross-lingual transfer abilities of the resulting models because they have been specialized for a single language. In this paper, we perform multilingual adaptive fine-tuning (MAFT) on 17 most-resourced African languages and three other high-resource languages widely spoken on the African continent -- English, French, and Arabic to encourage cross-lingual transfer learning. Additionally, to further specialize the multilingual PLM, we removed vocabulary tokens from the embedding layer that corresponds to non-African writing scripts before MAFT, thus reducing the model size by around 50\%. Our evaluation on two multilingual PLMs (AfriBERTa and XLM-R) and three NLP tasks (NER, news topic classification, and sentiment classification) shows that our approach is competitive to applying LAFT on individual languages while requiring significantly less disk space. Finally, we show that our adapted PLM also improves the zero-shot cross-lingual transfer abilities of parameter efficient fine-tuning methods.

本文通过多语言自适应微调方法，改进了前馈神经网络(AfriBERTa和XLM-R)模型，使其适配17种最常用的非洲语言及其他3种在非洲大陆广泛使用的高资源语言，在NER、新闻主题分类和情感分类等自然语言处理任务中具有竞争性，且需要的磁盘空间明显少于单个语言的自适应微调方法。同时，本文的适配模型还能够改善参数效率微调方法的零-shot跨语言转移能力。

通过多语言自适应微调将预训练语言模型改进到非洲语言