BanglaNLG and BanglaT5: Benchmarks and Resources for Evaluating Low-Resource Natural Language Generation in Bangla
It addresses the problem of low-resource NLG for Bangla speakers and researchers, providing benchmarks and a model to advance future work, though it is incremental in applying existing methods to a new language.
This work tackles the lack of resources for natural language generation (NLG) in Bangla by introducing BanglaNLG, a benchmark with six tasks including a new dialogue dataset, and BanglaT5, a pretrained Transformer model that achieves state-of-the-art performance with up to 9% absolute and 32% relative gains over multilingual models.
This work presents BanglaNLG, a comprehensive benchmark for evaluating natural language generation (NLG) models in Bangla, a widely spoken yet low-resource language. We aggregate six challenging conditional text generation tasks under the BanglaNLG benchmark, introducing a new dataset on dialogue generation in the process. Furthermore, using a clean corpus of 27.5 GB of Bangla data, we pretrain BanglaT5, a sequence-to-sequence Transformer language model for Bangla. BanglaT5 achieves state-of-the-art performance in all of these tasks, outperforming several multilingual models by up to 9% absolute gain and 32% relative gain. We are making the new dialogue dataset and the BanglaT5 model publicly available at https://github.com/csebuetnlp/BanglaNLG in the hope of advancing future research on Bangla NLG.