Abstract
The use of artificial intelligence (AI) has significantly streamlined a range of academic tasks, notably in terms of quality and efficiency. The study examines both the quality and efficiency of AI-driven paraphrasing tools in preserving the originality of the source text. We employ a mixed-method approach of qualitative and quantitative analyses based on Taylor’s (2017) lexical semantic framework and previous research on AI paraphrasing strategies. The study focuses on the paraphrased output of nouns generated by the top AI tools used for academic writing: QuillBot and Paraphraser. The corpus of 100 abstracts (hard science and soft science) was collected from the ScienceDirect database. The analysis examines two categories of AI paraphrasing strategies: structure-related and lexical semantic-related. A jury of linguists assessed the researchers’ strategy classification for validity and accuracy. The findings show that while QuillBot scores the top in preserving meaning in hard science abstracts, Paraphraser scores the lowest due to its distortion of the grammatical structure of the source text. For soft science abstracts, the Paraphraser-AI tool produces similar output, generalising concepts in soft science abstracts, hence altering the original meaning. It also reveals that the top paraphrasing strategies used among the two AI tools were using synonyms and changing parts of speech. However, QuillBot used fewer synonyms in hard science than in soft science abstracts, therefore preserving technical terms, clarity, and quality of the original text. Pedagogically, the findings serve as a guideline for scholarly writing by replicating the most accurate AI paraphrasing strategies in preserving the meaning.
Presenters
Rima Jamil MalkawiStudent, PhD Candidate in Linguistics and Translation, University of Sharjah, United Arab Emirates
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
KEYWORDS
Lexical semantics; Artificial intelligence; Paraphrasing tools; Paraphrasing; Education
