Is UDA(for text classification) purely based on back-translation?

[abrhaleitela]

I wanted to augment labeled data for sentiment analysis of low resource language where there are no reasonable MT models to do back-translation. Is there a way to do that without using back-translation?

Thanks

[michaelpulsewidth]

You can also try a simpler data augmentation SwitchOut, which replace words with words randomly sampled from your vocabulary.