ParaGen: Paraphrase Generation Using NLP
1) P. Sejal 2) A. Shanthi Priya 3) Gaurav Tripathi 4) N. Sohanmani 5) D. Sreeja Reddy
Guide: Prof K Divya Bharathi
Department of Artificial Intelligence and Machine Learning (AI&ML)
Malla Reddy University, Maisammaguda, Hyderabad
2111CS020501@mallareddyuniversity.ac.in, 2111CS020503@mallareddyuniversity.ac.in
2111CS020529@mallareddyuniversity.ac.in , 2111CS020530@mallareddyuniversity.ac.in
2111CS020540@mallareddyuniversity.ac.in, k.divyabharathi@mallareddyuniversity.ac.in
ABSTRACT
This project focuses on paraphrase generation using various datasets and advanced transformer models. The datasets include Quora Question Pairs, the Microsoft Research Paraphrase Corpus (MRPC), and MS COCO Annotations. The Quora dataset involves pairs of questions to determine semantic equivalence. MRPC comprises sentence pairs extracted from online news sources, annotated for semantic similarity. MS COCO provides captions for images, with each image having multiple paraphrased captions, which are broken into pairs for analysis. The project explores the use of transformer models such as T5 (Text-to-Text Transfer Transformer), BART (Bidirectional and Auto-Regressive Transformers), and a combination of transformer models with sequence-to-sequence models. Additionally, it investigates the encoder decoder model using BERT as both encoder and decoder, implemented through the Hugging Face library. These models are fine-tuned for specific tasks by modifying input and output sequences to include task names. The goal is to achieve state-of-the-art results in paraphrase generation, enhancing natural language processing applications. By leveraging these diverse datasets and sophisticated models, the project aims to improve the understanding and generation of semantically equivalent text, thereby advancing +the field of NLP and its practical applications in various domains such as machine translation, information retrieval, and question-answering systems.
