⚠️ working prototype — results are not fully verified yet · lp2graph repo ↗
Conference Demo · LP Mining

LP Mining with LP2Graph

Mining published LP/MILP formulations for railway rescheduling into a reproducible, typed-graph dataset — corpus construction & human-in-the-loop review.
corpus snapshot 2026-06-21 · 238 papers · 8,957 formulas

PRISMA — corpus pipeline

📊 PRISMA corpus yield
Open ›
The corpus-acquisition pipeline as a PRISMA-2020 flow: identification (database + citation snowballing) → screening → retrieval & eligibility → included, with per-stage yield, losses and amplification.
10,099 records in238 included×37.6 formula yield

Corpus at a glance

10,099
records identified
6 database queries + citation snowballing
286
reports retrieved
43 database · 243 snowball
238
source papers included
83.2% eligibility yield
8,957
candidate formulations
MathML → LaTeX, Tier-2
Why only 3 papers in the demo? The public demo ships only papers published under a CC-BY licence (verified via OpenAlex), and is deliberately focused on a small sample of simple, single-objective MILP formulations. The full 238-paper corpus was mined under Elsevier text-and-data-mining terms and is reproduced from the pipeline, not redistributed.

Formula Express — review game

🚆 Formula Express — corpus review
Play ›
The gamified human-in-the-loop review app on an exemplary subset of the corpus: per-paper symbol graphs, formula expression trees, accept ✓ / fix ✎ / reject ✗, P1–P5 classification, XP & streaks. Works offline once loaded; progress stays on your device.
3 papers92 formulassingle-objective MILPsall CC-BY open access
⏱️ Blitz 🧭 Shell Sorter 📔 Journal 📚 Papers

Your review progress

0
🔥 day streak
0
XP earned
0
decisions today
0%
formulas reviewed
Overall progress
sister site
🕸 LP2Graph — typed graphs & configurator
Open ›
Accepted formulas become canonical typed-graph Formulations with provenance — the dataset behind the paper's taxonomy. Explore the representation and the model configurator.