Plagiarism Detection Tool

Overview

A web-based plagiarism detection system developed during a software development internship at Batunet Bilişim Hizmetleri. The system measures textual similarity across multiple documents, produces a percentage similarity score for every document pair and surfaces the overlapping passages.

The Need

Assessing document originality spans a wide spectrum, from verbatim copying to restructured text. Approaches built on a single algorithm fall short of capturing that range. The goal was a system that measures different kinds of similarity consistently, presents its findings in a clear report and is accessible from the browser.

The Solution

The system is built on a four-stage analysis pipeline whose stages complement one another:

1. Normalisation — Text is stripped of differences in case, punctuation, whitespace and Turkish characters, so the comparison responds to content rather than formatting.
2. Shingling — Documents are split into word n-grams, allowing even relocated sentences to be detected through matching fragments.
3. Multi-measure similarity — Jaccard similarity over shared fragments and cosine similarity over word-frequency vectors are computed together, capturing verbatim and restructured text in the same analysis.
4. Reporting — A similarity score for each document pair, with a side-by-side view highlighting matching passages.

Architecture and Web Migration

The system was first developed as a desktop Java application and then moved to the web with a JSP-based interface. Keeping the business logic in classes independent of the presentation layer made it possible to complete this migration without touching the analysis algorithms. Uploading, the results table and the comparison view are now fully available in the browser.

Impact

• An explainable analysis structure that brings different similarity types together in a single report.
• Fragments are generated once and reused, efficiently managing a comparison load that grows quadratically with the number of documents.
• A layered architecture gives the system a foundation that can be extended to new interfaces and use cases.
• The project was delivered within a professional team's coding standards and handover processes.