A Multimodal Assignment Plagiarism Detection System: A Case Study in Android Development Course

In Android application development courses, student assignments typically comprise multimodal content, including source code, interface layout files, and runtime UI screenshots in reports. This multimodal nature poses significant challenges for traditional text-based plagiarism detection methods in comprehensively evaluating assignment originality. As programming education continues to expand, assignment plagiarism has become increasingly prevalent, seriously impacting teaching quality and learning outcomes. To address this challenge, this paper presents a deep learning-based multimodal plagiarism detection system. The system employs ConvNeXt to generate image features, BERT (Bidirectional Encoder Representations from Transformers) to analyze textual descriptions and CodeBERT to analyze source code. These features are then concatenated and processed through a Multi-Layer Perceptron to generate a unique digital fingerprint for each assignment. The system identifies potential plagiarism by computing similarity scores between different features and the fingerprints. Experimental results in real educational settings demonstrate the system's effectiveness in detecting various forms of assignment plagiarism, significantly improving both accuracy and efficiency of plagiarism detection, thus providing a powerful tool for maintaining educational quality.