Why papers pile up in a mess
- File names carry no information. A PDF from a preprint server or a journal page is often named with an ID or a string of characters, and you cannot tell which paper it is without opening it.
- Sources are mixed. For one topic you collect from a university database, from colleagues, from conference pages, each with its own naming.
- Finding one depends on memory. You remember "the one about attention" and then preview files until you hit it.
How Filewise recognizes a paper
It does not go by file name. It reads the text layer of the PDF, and treats a file as a paper only when it finds one of three kinds of signal:
- An arXiv ID. The word arXiv together with an ID such as 2301.12345. The word is required so that version numbers or amounts are not mistaken for paper IDs.
- A DOI. A DOI in standard format in the text.
- Abstract plus references. An Abstract section together with References or a Bibliography, or with Keywords or Index Terms fields as in IEEE style.
The test is deliberately strict. A white paper or research report that only has an abstract and a conclusion is not treated as a paper. Missing a file is better than renaming the wrong one.
What it renames papers to
A name is built from three parts:
- First author's surname and a short title. The local model reads them from the text, in English for English papers. What it reads must be findable in the original text; if it is not, the answer is dropped and the name falls back to a generic "paper".
- The arXiv ID. When one is found it is appended, so you can get back to the source page.
- A date. The date near "Submitted" or "Published" in the text is preferred, with the file date as the fallback. Whether the date goes first or last and how it is written can be changed in Settings.
Illustration (real names depend on your files): 2017-06-12_Vaswani_attention-is-all-you-need_paper_arXiv-1706.03762.pdf
Where the file goes is up to you
For papers, Filewise only suggests a new name. It does not suggest a destination folder, because the same paper can sensibly live under a topic, a year or a project, and only you know which. After renaming you can move files into your own structure. If you have not settled on one, see How to organize files on a Mac.
What it does not do
- It is not a reference manager. It does not manage citations, fetch metadata or format bibliographies. If you already use a reference manager, keep using it; Filewise deals with the loose PDFs in Downloads and on the Desktop.
- It can misread. The local model sometimes mixes up the author or takes a running header for the title. That is why every rename is preceded by an "old name to new name" plan that you approve.
- Unusual formats may not be recognized. Technical reports without an abstract, book chapters and whole proceedings volumes may not meet any of the three signals and keep their names.
- Files without a text layer. A PDF with no readable text offers no signal, and it keeps its name as well.
A safe way to try it
- Pick a folder with a dozen or so papers and ask Filewise for a plan.
- Read the new names line by line, untick the wrong ones, and run the rest.
- Every step lands in an undo log that still works after a restart.
Requirements: Apple silicon, macOS 14 or later. Reading and naming happen on your Mac and file contents are not uploaded. Free to download.
FAQ
Are my papers uploaded to the cloud?
No. Reading, deciding and naming all run offline on your Mac, and file contents never leave the machine.
Does it work with papers in Chinese?
Yes. The signals include the Chinese words for abstract and references, and names are written in Chinese for Chinese papers.
Will it move papers into folders for me?
For papers it only suggests a new name and does not suggest a destination. Where a renamed paper goes is your decision.
