Has anyone considered computationally (e.g., Google site mining) identifying repeat patterns in publishing sites, building a site-similarity rank index, then letting millions of scholars judge for themselves? Most of these sites are clones of bulk predatory publishers (i.e., they self-plagiarize their sites, because doing otherwise is not as profitable as slumlording). Legitimate ones are far more unique.
I just Google-searched "Submission of a manuscript implies that authors have met the requirements of the editorial policy" (a string I selected from twasp.info/journal). Dozens (hundreds? thousands?) of sites popped up. There's no Jeff Beale to attack in such a model, because scholars are only confirming or rejecting given sites.
The algorithm would assume that the greater the difference between a given journal website and its peers, the greater the level of unique human curation and the greater likelihood of legitimacy. Certain small, arguably legitimate journals may fall close to this "questionable" category, but conceivably they will have (or will develop) certain distinctive features. Has anyone attempted this?
I don't say that this kind of work is not without its problems. Some spring to mind:
- What then is "legitimate"? Elsevier's tens of thousands of high-quality journal sites may arguably be even more homogeneous than a bulk predatory aggregator and so could well fall into the questionable category. (But then presumably certain specifics, such as reviewing rules and other fine details, that are unique from beginning to end.)
- Diligent publishers could learn how to hack around this problem by creating their own "uniqueness engines" to add noise between their sites. (But then that's a big job to do right, isn't it?)
- Something like this couldn't work pure out of the box. There would need to be some "shims" in place -- such as, perhaps, strong weighted scoring for the presence of certain key signs of legitimate sites (a real e-mail address? some AI to find a real and verifiable person as the editor-in-chief?). And even the algorithm couldn't simply be a "difference engine": it would need to be highly tuned for the given task, such as my example at top, which would weight high for certain details with the scent of true legitimacy, and ignore certain other details. The algorithm would only be as good as the initial analysis of where and how to weight. And it would need to be adaptable to changing strategies over time.
Dear Peter Zelchenko , thanks for your kind letter. I have realized now that you have posted a fine question.
The example of https://researchleap.com/ and his editor from Croatia, Bojan Obrenovic, is typical example of bad publishing practices, predatory, of course.
https://www.linkedin.com/in/obrenovic-bojan/?originalSubdomain=hr
This editor also has a profile at RG.
Bojan Obrenovic
- Badges
- Science topic
- Similar topics
- Publishing