AMMUS: A Survey of Transformer-based Pretrained Models in Natural Language Processing

Transformer-based pretrained language models (T-PTLMs) have achieved great in almost every NLP task. The evolution of these models started with and BERT. These models are built on the top of transformers, -supervised learning and transfer learning. Transformed-based PTLMs learn language representations from large volumes of text data using-supervised learning and transfer this knowledge to downstream tasks. These provide good background knowledge to downstream tasks which avoids of downstream models from scratch. In this comprehensive survey paper, initially give a brief overview of self-supervised learning. Next, we various core concepts like pretraining, pretraining methods, tasks, embeddings and downstream adaptation methods. Next, we a new taxonomy of T-PTLMs and then give brief overview of various including both intrinsic and extrinsic. We present a summary of useful libraries to work with T-PTLMs. Finally, we highlight some of future research directions which will further improve these models. We believe that this comprehensive survey paper will serve as a good to learn the core concepts as well as to stay updated with the recent in T-PTLMs.