LINGUISTIC PERFORMANCE OF AUTOMATED PARAPHRASING TOOLS: A COMPARATIVE STUDY

Abstract

Recent years witnessed drastic change in automated paraphrasing tools that are greatly used by researchers, students and content writers required for their writing. The accuracy of these paraphrasing tools is outstanding but somehow linguistic efficiency of these tools is a questionable phenomenon. Thus, current study endeavored to identify and highlight linguistic flaws of these automated paraphrasing tools for further improvement. For this purpose, text from Donald Trump inauguration speech was selected as data. The collected data in its original textual form was processed through QuillBot, Spinbot, ChatGPT and ZeroGPT that was analyzed through linguistic theory of artificial flaws by Mehmood (2024) as well as by using Voyant, a corpus tool. The study found semantic differences, passive syntactic construction, irrelevant context, and violation of brevity in paraphrased text. Moreover, it was also found that despite of having need for further upgradation, paraphrasing performance of QuillBot is outstanding, paraphrasing technique of ChatGPT is satisfactory, performance of Spinbot is ordinary whereas ZeroGPT still needs greater work for better linguistic performance in paraphrasing.