A Similarity Detection Tool for Jupyter Notebooks

A Tool for Detecting Similarities in Jupyter Notebooks Used as Assessment Reports [Plagiarism Detection]

Learn more [Paper]

Similarity Score

...

Overall similarity score स_𝐴←𝐵 indicates fraction of contents in A similar to B.

• स ∈ [𝟎, 𝟏]
• Asymmetric: स_𝐴←𝐵 ≠ स_𝐵←𝐴

See Details ...

Adjusting the aggressiveness of similarity detection algorithms by tuning:

• Threshold 𝑇
• minimum block length 𝐿_𝑚

By adujsting 𝑇 and 𝐿_𝑚, JBEval can be adopted to diverse context, which allows us to use JBEval for assessments in different courses.

Adjusting & Adopting

...

Overall similarity score स_𝐴←𝐵 indicates fraction of contents in A similar to B.

See Details ...

Features

JBEval comes with following features:

Supports

Code/Text

JBEval detect similarity not only in code but also in text written in Jupyter Notebooks as part of report.

Currently it supports two programing languages and text contents written in Raw and Markdown cells

Code:Python
Code:R
Text:plain-text (Raw)
Text:HTML (Markdown)

Control aggressiveness

Control & Adopt

JBEval can be adopted to divers sets of context, by adjusting two main parameters.

Adjusting Threshold T and minimum block length L_min

Programming heavy reports (Computer Science)
Programing + Text (Physics/Chemistry)
Light Programing + Text (Humanities)
Project reports / Assignments

Validation with visualisation

Visualisations

JBEval produces amazing diagrams and pictures to visualise the similarity between two reports, which helps to validate the plagiarims.

Validating the similarity cases

Similarity Maps
Visualising entire cohorts report
helps to validate the plagiarism cases

Processing

Batch Processings

JBEval can be used to process batches of Jupyter-Notebooks, allowing to detect plagiarism with different cohorts

Processing in bulk

Fast batch processing
Comparing with previous cohorts
Detects plagiarism and collusions
Identifying groups of similar reports