Enhancing Plagiarism Detection in Marathi with a Weighted Ensemble of TF-IDF and BERT Embeddings for Low-Resource Language Processing

TL;DR

This paper proposes a novel ensemble method combining TF-IDF and BERT embeddings to improve plagiarism detection accuracy in Marathi, a low-resource language, addressing the gap in semantic analysis tools for regional languages.

Contribution

It introduces a weighted ensemble approach that integrates statistical and semantic text features specifically tailored for Marathi plagiarism detection, a low-resource language.

Findings

Enhanced detection accuracy over baseline models

Effective combination of TF-IDF and BERT embeddings

Addresses low-resource language challenges

Abstract

Plagiarism involves using another person's work or concepts without proper attribution, presenting them as original creations. With the growing amount of data communicated in regional languages such as Marathi -- one of India's regional languages -- it is crucial to design robust plagiarism detection systems tailored for low-resource languages. Language models like Bidirectional Encoder Representations from Transformers (BERT) have demonstrated exceptional capability in text representation and feature extraction, making them essential tools for semantic analysis and plagiarism detection. However, the application of BERT for low-resource languages remains under-explored, particularly in the context of plagiarism detection. This paper presents a method to enhance the accuracy of plagiarism detection for Marathi texts using BERT sentence embeddings in conjunction with Term Frequency-Inverse Document Frequency (TF-IDF) feature representation. This approach effectively captures statistical, semantic, and syntactic aspects of text features through a weighted voting ensemble of machine learning models.