Cross-Language Translation Algorithm Based on Word Vector and Syntactic Analysis

Abstract

Digital technology has created a necessity to protect intellectual uniqueness in translated works. Cross-lingual syntactic connection is critical for measuring the level of similarity between textual pairs produced in different languages for the purpose to detect plagiarism. This study addresses cross-lingual syntactic analysis and plagiarism detection and evaluation using the university student’s dataset. The data is utilized for translation and plagiarism identification, and it is transferred into a research model using the Natural Language Process (NLP). The data is then extracted into features using a word vector. The research proposed that the word Embedded Fast Recurrent Network (WE-FRN) is used for syntactic analysis and plagiarism detection. To handle the plag detection issue, many neural network designs were proposed, including regime proposal (plagiarism or independently created) and binary classification (syntactic regression analysis of documents). Experimental results indicated that utilizing WE-FRN with rich syntactic characteristics produced better outcomes than baseline and the loss function of the classification is also analyzed by the regression of the source and suspicious documents.