Few parents, many childs

Hi. We have an index containing documents that have a title field and each document logically belongs to a category.

Now we'd like to explore the option of taking a title and based on this title recommending the category to place this item in.

One way i thought of to make this work is to create a document for each category and then making the items child documents of a category

That way i would be able to do a has_child query, that would return the average score per category

However there are aprox 2000 categories and 10 million items. And i am wondering if I will run into any issues with this. Since it does not strike me as the typical usecase for parent child documents. Also if someone has a better idea to accomplish thr same goal i am open to suggestions