Financial News Summarisation using Transformer Neural Network

Abstract Transformer architecture, which is based on self-attention mechanism, revolutionised the field of NLP in 2017. It overcame many of the limitations of sequential and iterative approach of the previous popular architectures like LSTM. Generative Pretrained Transformer (GPT), which is a type of transformer model, is one of the most powerful neural network architectures for the purpose of text summarisation. This paper elaborates how and why GPT-2 can be used for financial news summarisation.