QUORA QUESTION PAIRS ANALYSIS USING PERT

Abstract

An innovative design called The Transformer in NLP tries to tackle sequence-to sequence problems while skillfully managing long-range relationships. It doesn’t use convolution or sequence-aligned RNNs; it just uses self-attention to compute repre sentations of its input and output. The encoder-decoder design is the foundation of the transformer concept. After conducting in-depth study, researchers put forward the BERT and GPT transformer-based models, which significantly improved the bulk of NLP tasks including text creation, text summarization, and question an swering, among others. But as time went on, a number of these models’ drawbacks became apparent. PERT was recommended as a way to get around one of these drawbacks. In this project work, we fine-tune the pre-trained model on the similarity and paraphrasing task and analyze how the model performs in comparison to the other previously introduced methods.