Rejected Article
Journal Recommendation via AI
Academic Journal Publication directs the development of knowledge for both academia and professional organizations. It is the backbone of global development and approximately more than 2.5 million papers are published each year in various disciplines from humanities to physical sciences. But even in this expansive and albeit strict field, there can be many detractors found - people who write plagiarized articles, people who alter their findings to fit in a popular field, who use flawed editing procedures and so on and so forth.
Our model proposes to rectify these problems and solve the problem of redundancy, and journal rejections due to misfit. The model offers a range of recommendations for alternative journals (author manuscript-to-journal matching using AI) to submit author's
article/paper.
American multinational publishing company
Client
ePublishing Giant
CLIENT
Our client is a content publishing house that operates globally and among various disciplines and specialties. They publish more than 1500+ journals, 200+ reference books, and they also publish standalone articles and books numbering in the thousands. Under these multiple Journals, they publish multiple articles every year, and each Journal may have many categories and sub-categories in it. In the current process, the author requests our client to publish their article in a specific category/sub-category in a Journal, and if the article fails to meet all the required criteria of that category/sub-category then the article gets rejected completely. With this they are also losing a chance of publishing the article in other Journals category/sub-categories, where it has a higher acceptance rate. So our client's expectations from us is, if they reject an author’s article in one Journal, then this article should be checked with another Journal's category/sub-category for any possibility to get accepted before rejecting it completely. Our model can easily identify patterns at each Journal level, and to its category and sub-category level to find the categorical semantic similarity, and match the article to the top possible categories/sub-categories.
CHALLENGE
A versatile model like this certainly has its own set of challenges, they are stated as below:
-
Title/Pseudonym Resemblance - The title of various articles can have word/lexicon level similar (but semantically different), and this can cause a major issue not just for cataloguing but also for differentiating between fraudulent and legitimate authors.
-
Intensive Data - Since our client is a multidisciplinary publishing house; the amount of data present can be excessive in both size and differentiation.
-
Document Closeness Issues - Sometimes, there could be multiple articles pertaining to the same type of content already wrongly falling in multiple Journals and category/sub-categories. This could cause issues regarding selective identification.
-
past rejection criteria are not so regulated - Due to no fixed rejection rules, and an informal and ironic nature of feedback process against each rejections, makes the challenge more critical.
SOLUTION
Solution summary:
-
Our model is a concise one, it can delineate the nature of previously rejected articles with the category they are published in and create pattern sets based on them.
-
The model built from these pattern sets can be processed and cross-referenced to the current articles, which can easily identify their occurrences of rejected articles in other journals as well.
-
This helps the article to get matched with other successful journal publications and helps in identifying a best fit Journal category/sub-category for the article publication.
-
This also prevents the article recursion in journal publication and we can easily identify the repeat offenders.
With these challenges in mind we have drafted a concrete solution to this problem:
-
A sample set is created by organizing the content of rejected and selected articles from the publisher, they are then organized with categories and sub-categories.
-
Various ML classification methodologies are applied to differentiate categories and sub-categories.
-
Newly received articles are then compared and contrasted with the original golden dataset.
-
Top five Journal category/sub-category are recommended which have higher acceptance rate are pushed for the publisher’s perusal.
RESULT
Implementation of this model provides many valuable developments for our clients. The most important goal achieved by this proposition is safeguarding the safety of science and other schools of thought, upholding consistency and truth. We can also increase the efficiency of publishing and prevent the wastage of resources on the publisher’s side. More legitimate scholars will be inclined to write for our publisher because of its increased credibility via implementing this data model.