Page 507 - Emerging Trends and Innovations in Web-Based Applications and Technologies

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470

Fig 3. Methodology of plagiarism detection tool
VI. PERFORMANCE EVALUATION 2. Training: Originality Guard was trained on the training
To comprehensively evaluate the performance of Originality set using a 5-fold cross-validation approach
Guard, a multi-faceted evaluation framework was designed,
incorporating various metrics, baselines, and experimental 3. Testing: the trained model was evaluated on the testing
setups. set
4. Baselines: two baseline methods were used for
Evaluation Metrics comparison:
The following metrics were employed to assess Originality  Turnitin: a commercial plagiarism detection tool
Guard's performance:  PLAGUE: a state-of-the-art machine learning-based
1. Precision: measures the proportion of true positives approach
among all detected plagiarism instances
VII. CONCLUSION
2. Recall: measures the proportion of true positives among
This review paper presented Originality Guard, a novel
all actual plagiarism instances
plagiarism detection tool that leverages advanced natural
3. F1-score: measures the harmonic mean of precision and language processing and machine learning techniques. The
recall main contributions of this research includes:
4. Accuracy: measures the proportion of correctly 1. Development of Originality Guard: A robust and
classified instances accurate plagiarism detection tool that outperforms
existing baselines.
5. Mean Average Precision (MAP): measures the average
precision at different recall levels 2. Comprehensive evaluation framework: A thorough
evaluation methodology that assesses the performance
6. Receiver Operating Characteristic (ROC) Curve: plots the
of Originality Guard using various metrics.
true positive rate against the false positive rate at
different thresholds 3. Insights into plagiarism detection: A detailed analysis of
the results, highlighting the strengths and weaknesses of
Experimental Setup
Originality Guard.
The experimental setup consisted of:
1. Dataset: a large-scale dataset of 50,000 academic It has significant implications for advancing plagiarism
documents, divided into training (80%) and testing detection and accuracy, promoting academic integrity, and
(20%) sets

IJTSRD | Special Issue on Emerging Trends and Innovations in Web-Based Applications and Technologies Page 497