stopslop

Measure a paper's prose against how a venue actually writes — before and after LLMs existed.

Two baselines, always both: fossil (published up to 2022, so it cannot have been machine-assisted) and current (the last full year plus this one). The fossil window is the target; the current window is context, never an excuse. See UNSLOP.md.

Install

The wheel carries the baselines for all seven venues — 105 venue-years of measurements — so there is nothing to crawl and nothing to configure.

pip install stopslop              # pypdf, BSD-3-Clause, no copyleft
pip install "stopslop[pymupdf]"   # adds PyMuPDF (AGPL-3.0, opt-in)
pip install -e .                  # editable, for hacking on it

Then, from any directory, in any project:

stopslop check paper.pdf --venue tmlr --out report.md

No system tools required. Everything else is stdlib re and statistics.

Extractors

The extractor is part of the measurement. Across 24 papers the four supported extractors agree to within ~1% on prose metrics — first person 18.1–18.3, -ly adverbs 13.2–13.4, passives 10.3–10.5 — and disagree sharply on layout-sensitive ones: numerals runs 43.4 under pymupdf against 61.4 under pypdf. So measurements are stored per extractor and never mixed, and stopslop check refuses rather than compare across them.

Baselines for all of them ship in the base wheel, so an extra only decides what you can measure with, never what you can read. Pick one with --extractor or $UNSLOP_EXTRACTOR.

pip install stopslop carries no copyleft: PyMuPDF is AGPL and is opt-in, so taking on that licence is always your deliberate choice.

Use

stopslop check paper.pdf --venue tmlr --out reports/paper.md   # score one paper
stopslop trend tmlr --out reports/tmlr_trend.md                # one venue's drift
stopslop crosstrend --out reports/total_trend_points.md        # venues kept apart
stopslop pooltrend  --out reports/total_trend.md               # venues pooled
stopslop stats tmlr 2022                                       # one crawled year
stopslop crawl tmlr 2022 2025 2026 --limit 20                  # fetch and measure

crawl is the only command that touches the network, and it needs no key or account. Everything else is offline.

Where the baselines come from

Resolved at call time, most specific first:

1. --root DIR on any command
2. $UNSLOP_MEASUREMENTS
3. ./data in the working directory, if it exists
4. the measurements bundled in the package

So a project that measures its own papers into ./data uses them automatically, and everyone else gets the shipped ones. PDFs are cached separately, under --pdf-root / $UNSLOP_PDFS / ./corpus.

Plots

--plot needs matplotlib, which is not a dependency and not an extra. Plotting is an analytics concern and the measurement path has no use for it, so it goes in a separate environment:

python3 -m venv .venv-analytics
.venv-analytics/bin/pip install -e . -r analytics-requirements.txt
.venv-analytics/bin/stopslop trend tmlr --plot --out reports/tmlr_trend.md

Every other command works without it.

Development

pip install -e ".[pymupdf,pypdfium2,pdfminer]" --group dev
pytest
stopslop remeasure --extractor pypdf --root src/stopslop/data   # offline, no network

Measurements are generated artifacts and are written straight into the package, so there is no snapshot to keep in sync. corpus/ is a pure PDF cache: large, re-fetchable, and entirely gitignored. remeasure re-reads those PDFs under a chosen extractor without touching the network — crawl would re-query arXiv for a paper list we already have.

Layout

src/stopslop/
  extractors/        pypdf | pymupdf | pypdfium2 | pdfminer
  data/<extractor>/<venue>/<year>/stats.json
                     the measurements, shipped in the wheel
corpus/<venue>/<year>/*.pdf
                     PDF cache: 3.7 GB, gitignored, re-fetchable
reports/             generated trend reports and checked papers
tests/

Status

Pre-release. Published to PyPI as stopslop: the obvious names, unslop and deslop, are both taken by humanizers, which do the opposite of what this tool is for --- it measures and locates, it never edits your paper. Every candidate that was checked is in "The rename" in UNSLOP.md, and the remaining release blockers are in the Decision section of COMPETITION.md.

MIT licensed.