Semi-Supervised Low-Resource Style Transfer of Indonesian Informal to Formal Language with Iterative Forward-Translation

In its daily use, the Indonesian language is riddled with informality, that\nis, deviations from the standard in terms of vocabulary, spelling, and word\norder. On the other hand, current available Indonesian NLP models are typically\ndeveloped with the standard Indonesian in mind. In this work, we address a\nstyle-transfer from informal to formal Indonesian as a low-resource machine\ntranslation problem. We build a new dataset of parallel sentences of informal\nIndonesian and its formal counterpart. We benchmark several strategies to\nperform style transfer from informal to formal Indonesian. We also explore\naugmenting the training set with artificial forward-translated data. Since we\nare dealing with an extremely low-resource setting, we find that a phrase-based\nmachine translation approach outperforms the Transformer-based approach.\nAlternatively, a pre-trained GPT-2 fined-tuned to this task performed equally\nwell but costs more computational resource. Our findings show a promising step\ntowards leveraging machine translation models for style transfer. Our code and\ndata are available in https://github.com/haryoa/stif-indonesia\n