Compilatio Magister and Magister+ analyses: several indicators
In your analysis report Compilatio Magister or Compilatio Magister+ you have access to several indicators:
- Overall percentage of suspicious texts
- Semantic similarity score (deep paraphrasing)
- Multilingual similarity score: side-by-side comparison between the analyzed text and the source text in the original language*
- Score of texts potentially generated by AI*
- Score for unrecognized languages (potentially altered texts)
- Grammar & spelling checker*
*To consult this score, you need a subscription to Compilatio Magister+
Why isn't the overall percentage of suspicious texts equal to the sum of the three scores?
It should be noted that the overall percentage of suspicious texts does not correspond to the sum of the three scores: similarity score, score of potentially AI-generated texts and score of unrecognized languages.
Suspicious passages can sometimes overlap.
A passage containing similarities with a web page, for example, can also be detected as a potentially generated by AI. The same passage may have languages not recognized.
Illustration with a case study
Visit a concrete example where we can identify overlaps between different types of suspicious passages:
A student uses identical Wikipedia content in his assignment.
As Wikipedia is a collaborative encyclopedia where everyone is free to add whatever content they wish, the source of the content is not verified.
Let's assume that the Wikipedia content retrieved was originally created by a generative AI. In the analysis report, the passage copied and pasted by the student will be considered as potentially generated by AI. At the same time, similarities with the Wikipedia source will be highlighted.
The overall percentage of suspicious texts will not equal the total of the 3 scores: similarity score, score of texts potentially generated by AI, score of unrecognized languages.
Here is a visual representation of the distribution of these 3 scores in the text:
📌 Questions about Magister, Magister+, plagiarism or AI?
Get answers live during our "Q&A Webinar".
👉 Register for the next session
This article has been automatically translated. If you notice a translation error, please contact us.