An algorithm that isolates statistically improbable phrases. Because phrases are figments of thoughts, we are not so concerned with the grammar as we are with the logic or sensibility of the text content. It is also more robust with respect to translation flaws and artifacts.

https://en.m.wikipedia.org/wiki/Statistically_improbable_phrase

Amazon uses it to tag what is unique to a book in terms of its content. It uses a proprietary algorithm. Some common ones that may isolate statistically improbable phrases:

https://www.researchgate.net/publication/44601755_Identifying_duplicate_content_using_Statistically_Improbable_Phrases

Moving on to 'socially' improbable phrases as a means to lump and cluster content

https://www.researchgate.net/publication/253189906_Exploratory_Community_Sensing_in_Social_Networks

I also recommend Wolfram Alpha's inbuilt text analysis functions, when used in combination, like a triangulation, may allow for authenticity assessments:

https://www.wolframalpha.com/examples/NotableTexts.html

Perhaps the algorithm that you envisage can be decomposed into these?

Article Identifying duplicate content using Statistically Improbable Phrases

Article Exploratory Community Sensing in Social Networks

 Dear Shian-Loong,

thanks a lot for a detailed response. But I'm sorry for my intransparent way of formulating my question. I am concerned about authenticity when an interpreter translates a source text into another target language for it is his/her job. Not the way one wants to "rent" another person's ideas.

I am sorry. Anyway I've learnt much interesting from other fields.

Hi, Brahim! Thank you very much.