Methods in Low Resource settings can be inspired from other areas like Machine Learning and Computer Vision. Some approaches are the following.

Meta-learning: One can identify suitable auxiliary task for which there is a lot of training data available and use the trained model for transfer learning to complete the target task.

Adversarial Training: This training decreases differences in features between the pre-training and the target domain. Therefore, preventing the model from learning a feature-representation specific to a data source. This leads to creation of language independent representations which is a step closer creating a truly multilingual model.