Utilizing Deep Natural Language Processing to Detect Plagiarism

Abstract

Plagiarism is now pervasive in a variety of spheres of life, including academia and research. The development of plagiarism strategies used by plagiarists makes it difficult for existing approaches to accurately detect plagiarism. Plagiarism is checked using a variety of aspects, including syntactic, lexical, semantic, and structural features. This study examines novel and contemporary plagiarism detection tasks, particularly text-based and monolingual plagiarism detection. We suggested a four-stage innovative approach for detecting plagiarism. The natural language processing (NLP) methodology is used in this framework as opposed to the more conventional string-matching methods. By combining two metrics—skip gram and dice coefficient—on the basis of a corpus-based approach, this system investigates text similarity. Using the deep and shallow NLP approach, the text's deeper meaning is investigated. Our findings indicate that deep NLP is swiftly recognizing heavy revision. Shallow NLP efficiently prepares text for future processing. The findings of Word2vec are comparable to those of straightforward deep NLP techniques, however Word2vec also emphasizes documents that other methods might miss. Deep NLP also records changes in synonyms and phrases.