Manage research PDFs without losing the argument

A postgraduate folder can reach a thousand papers before it feels like a library. At that size, a reference manager is useful, but it is not enough on its own. Its database can break, its cloud sync can duplicate records, and its export format may not preserve the small decisions you made while reading. The PDF files and your notes need to make sense even when opened outside that app.

The aim is not to build a perfect archive. It is to stop losing time to the same mechanical mistakes: opening the same article five times, copying a quotation without its page number, or discovering too late that a scanned paper has no searchable text. A durable workflow starts with boring file names and ends with notes you can cite without guessing.

Give every paper a name that survives the database

Use one convention everywhere:

author-year-keyword.pdf

For example, nguyen-2021-rural-water-governance.pdf is much more useful than ScienceDirect-1-s2.0-S004896972034.pdf. Put the first author's surname first, then the publication year, then one phrase that says why the paper is in your pile. If two papers collide, add a second keyword rather than changing the pattern.

Folder structure should be just as plain. One workable version is:

• 00-inbox for new papers not yet triaged.
• 10-reading for papers you intend to read properly.
• 20-cited for papers already used in a draft.
• 90-discarded for papers you checked and decided not to use.

Do not delete a discarded paper immediately unless storage is a problem. Moving it to 90-discarded records the decision and prevents another afternoon spent rediscovering why it was not useful. A reference manager can still hold the citation metadata, but the file system should remain intelligible without it.

Read in passes, not in one heroic sitting

A long paper does not deserve a full read just because it has been downloaded. Use passes.

First pass: title, abstract, introduction ending, figures, tables, and conclusion. Record one of three decisions in a reading log: discard, skim later, or full read. A plain Markdown table is enough:

File	Decision	Reason	Date
nguyen-2021-rural-water-governance.pdf	full read	Method matches case study design	2026-08-10

Second pass: method and evidence. This is where you check whether the article can support the claim you might cite it for. Is the sample size relevant? Are the dates and geography comparable? Does the paper measure the thing your draft will say it measures?

Third pass: full read, but only for papers that survived the first two passes. By then you know what you are reading for, so your notes become sharper. You are not summarising the entire article; you are extracting the parts that can shape your own work.

A browser-local summarise PDF tool can help at the triage stage by giving you a rough map of a long document. Treat that map as a queueing aid, not evidence. If a paper will appear in your bibliography, read the relevant pages yourself and check the source before citing it.

Keep literature notes separate from permanent notes

A literature note belongs to the paper. It says what the source argues, what it measured, which quotation might matter, and where the evidence sits. A permanent note belongs to your argument. It states an idea in your own words and connects it to other ideas you are developing.

That separation prevents copied material from masquerading as thinking. In a literature note, use page numbers aggressively:

• p. 14: Authors define service reliability as weekly continuity, not daily pressure.
• p. 22 quotation: "..."

Do not trust memory for page numbers. Attach the page at the moment you copy or paraphrase. Later, when a supervisor asks whether a sentence is supported, you can open the paper and check the exact page.

Plain text or Markdown is deliberately unglamorous here. A .md file can be searched, backed up, diffed, and converted later. If you want to share a reading memo, the Markdown to PDF tool gives you a stable PDF without locking your notes into one note-taking app.

Extract text carefully from awkward PDFs

Many academic PDFs are two-column documents. A naive copy-paste can interleave the left and right columns, turning two good paragraphs into one unusable string. If you see sentences jumping across the page, stop and use a text extraction route that understands reading order.

The extract text tool is useful when the PDF already has a text layer. After extraction, check three things before pasting into notes:

1. Column order follows the article rather than the visual x-position alone.
2. Hyphenated line breaks have not created false words.
3. Footers, page numbers, and running headers have not been mixed into the paragraph.

Scanned or photographed papers need a different step first. A scan is just an image until OCR adds recognised text. Run the file through the OCR PDF tool before expecting search, selection, or extraction to work. Bad scans may still need manual correction, especially for equations, tables, marginal notes, and older typefaces.

A maintenance routine for a large pile

Once a week, empty 00-inbox. Rename every file, run OCR on scans, and make a triage decision. Once a month, search the reading log for skim later and either promote those papers to a full read or move them to 90-discarded.

The system works because every action leaves a trace outside any one app. The filename tells you what the paper is. The folder tells you its status. The reading log tells you why you kept or rejected it. The literature note preserves evidence with page numbers. The permanent note carries only the ideas that have earned a place in your own argument.

Checklist before a paper enters your thesis notes

• Rename the PDF using author-year-keyword.
• Record a triage decision so you do not reopen weak papers repeatedly.
• OCR scanned papers before searching or extracting text.
• Put every quotation beside a page number.
• Rewrite permanent notes in your own words and connect them to your argument.