Recent advances in the pre-training for language models leverage large-scale datasets to create multilingual models. However, low-resource languages are mostly left out in these datasets. This is primarily because many widely spoken languages that are not well represented on the web and therefore excluded from the large-scale crawls for datasets. Furthermore, downstream users of these models are restricted to the selection of languages originally chosen for pre-training. This work investigates how to optimally leverage existing pre-trained models to create low-resource translation systems for 16 African languages. We focus on two questions: 1) How can pre-trained models be used for languages not included in the initial pretraining? and 2) How can the resulting translation models effectively transfer to new domains? To answer these questions, we create a novel African news corpus covering 16 languages, of which eight languages are not part of any existing evaluation dataset. We demonstrate that the most effective strategy for transferring both additional languages and additional domains is to leverage small quantities of high-quality translation data to fine-tune large pre-trained models.

Identification of transliterations is aimed at enriching multilingual lexicons and improving performance in various Natural Language Processing (NLP) applications including Cross Language Information Retrieval (CLIR) and Machine Translation (MT). This paper describes work aimed at using the widely applied graphical models approach of ‘Dynamic Bayesian Networks (DBNs) to transliteration identification. The task of estimating transliteration similarity is not very different from specific identification tasks where DBNs have been successfully applied; it is also possible to adapt DBN models from the other identification domains to the transliteration identification domain. In particular, we investigate the applicability of a DBN framework initially proposed by Filali and Bilmes (2005) to learn edit distance estimation parameters for use in pronunciation classification. The DBN framework enables the specification of a variety of models representing different factors that can affect string similarity estimation. Three DBN models associated with two of the DBN classes originally specified by Filali and Bilmes (2005) have been tested on an experimental set up of Russian-English transliteration identification. Two of the DBN models result in high transliteration identification accuracy and combining the models leads to even much better transliteration identification accuracy.