Return to Answer

• Much faster than citation managers or search engines.
• Future-proof. Presumably computers will have some sort of hierarchical file system in the future. (Also see below about losing journal access being a reason to keep PDFs.)
• Uses the existing file system, which means no additional software is required. This also means there is no vendor lock-in. And you can use a wide variety of existing software to do local search, for example, using the native search tools, Everything, or Recoll.
• I can get hyperspecific on the topic. To give an example, dimensional analysis is an interest of mine. Most classifications might have only one subdivision for the subject. I have about 150.
• I personally find that organizing documents helps improve my understanding of a subject. It often takes a fair amount of knowledge to know how to categorize something.
• When compared against tagging, using a folder hierarchy is basically "tags with inheritance". This makes getting very precise easy as you can simply place a file deep in the hierarchy. To do the the same with tags, you'd have to apply more tags, which usually takes longer in my experience with software that uses tags.

• Much faster than citation managers or search engines.
• Future-proof. Presumably computers will have some sort of hierarchical file system in the future. (Also see below about losing journal access being a reason to keep PDFs.)
• Uses the existing file system, which means no additional software is required. This also means there is no vendor lock-in. And you can use a wide variety of existing software to do local search, for example, using the native search tools, Everything, or Recoll.
• I can get hyperspecific on the topic. To give an example, dimensional analysis is an interest of mine. Most classifications might have only one subdivision for the subject. I have about 150.
• I personally find that organizing documents helps improve my understanding of a subject. It often takes a fair amount of knowledge to know how to categorize something.
• When compared against tagging, using a folder hierarchy is basically "tags with inheritance". This makes getting very precise easy as you can simply place a file deep in the hierarchy. To do the the same with tags, you'd have to apply more tags, which usually takes longer in my experience with software that uses tags.

And academic search engines today focus heavily on text search. I know from my experience searching patents that classification search is essential to get around the problem of unknown synonyms in text search. For more esoteric topics like those studied by academics, I haven't found semantic search engines like Google Scholar to consistently understand what's a synonym, so that's not a replacement. One way to do classification search is to make the classification yourself, as I do.

And academic search engines today focus heavily on text search. I know from my experience searching patents that classification search is essential to get around the problem of unknown synonyms in text search. For more esoteric topics like those studied by academics, I haven't found semantic search engines like Google Scholar to consistently understand what's a synonym, so that's not a replacement. One way to do classification search is to make the classification yourself, as I do. (Another disadvantage of text search is information that doesn't appear in the text or is difficult to find with text search. This includes information in figures and anything that requires some sort of analysis to determine.)

I wrote more details about my system at r/datacurator.