Intelligent bar chart plagiarism detection in documents

Area of Science:

• Data mining
• Document analysis
• Computer vision

Background:

• Traditional optical character recognition (OCR) struggles with extracting data from complex documents containing graphical elements.
• Many documents with charts and text lack machine-readable data, limiting information retrieval.
• Plagiarism detection in visual data, such as bar charts, remains a challenge.

Purpose of the Study:

• To present a novel features mining approach for documents not amenable to OCR.
• To extract precise data values (Start, End, Exact) from bar charts by analyzing text-graphical relationships.
• To develop a robust method for detecting plagiarism in bar charts using textual and graphical features.

Main Methods:

• Analyzing the interplay between textual and graphical components within documents.
• Implementing a technique to extract Start, End, and Exact values from bar chart elements.
• Utilizing word 2-gram and Euclidean distance algorithms for plagiarism detection in bar charts.

Main Results:

• Successfully extracted key data points from bar charts in documents previously inaccessible to OCR.
• Demonstrated the capability to identify and quantify plagiarism within bar chart visualizations.
• Achieved accurate data extraction and plagiarism detection through the proposed integrated approach.

Conclusions:

• The novel features mining approach effectively overcomes OCR limitations for data extraction from graphical documents.
• The method provides a reliable solution for identifying and preventing plagiarism in visual data representations.
• This technique enhances information retrieval and integrity in document analysis.