Paraphrasing Tool Using Machine Learning

Abstract

Generating paraphrase has been a wide research area in the domain of natural language processing. Given an input statement, the paraphrase generating system will generate a new statement with the same intent but a different sentence structure or vocabulary. It has numerous implementations and advantages in the real world. It can be used for author anonymity applications, text summarizing applications, text augmentation, or plagiarism detection applications. In this study, our goal is to build a basic paraphrasing system using machine learning, where provided an input statement, the tool outputs a list of paraphrases for users to choose from. Our focus remains on using a sequence-to-sequence encoder-decoder-based model to develop this system. Once the model is deployed, it is integrated with the web user interface for anyone to use. This paper focuses on discussing the evolution of approaches used for paraphrase generation over time. It talks about the traditional approaches as well as the state-of-the-art neural approaches for generating paraphrases. Moreover, we present the experimental work conducted using one such neural approach, along with the result analysis and discussion for future work.