On-premise document parser
Turn HWP documents into RAG-ready data,
inside your own servers.
Luminir Document Parser is an on-premise document parser built by LEMONG. It finds titles, text, tables and figures in HWP·HWPX, PDF, Office and scanned documents and outputs them in reading order as Markdown and JSON. The recognition models are installed on your server together with the parser.
Contact usHow it works
Accept files
34 file types including HWP·HWPX, PDF, Word, PowerPoint, Excel and images, without converting them first.
Layout analysis
Detects titles, body text, tables, figures and page headers on every page, and decides the reading order.
Region recognition
Reads text as text, tables as rows and columns including merged cells, and formulas as LaTeX.
Output
Markdown plus JSON with the type, page number and coordinates of each element. Headers and page numbers are labeled so you can drop them easily.
Built for real work: Direct HWP·HWPX parsing · Heading levels · Header and page-number labels · Element coordinates · Batch conversion with resume · Closed-network deployment.
Public benchmarks (self-measured)
| Benchmark | Data | Score | Comparison | Date |
|---|---|---|---|---|
| opendataloader-bench | Published by Hancom · 200 English pages | 0.914 | Above the public #1 score (0.907) | 2026-10-12 |
| KoOCR-Bench | schift-io v1.2 · 623 Korean public-sector pages | 0.759 | Upper tier of the public leaderboard | 2026-10-12 |
| DP-Bench | Published by Upstage · 200 English pages · table TEDS | 0.962 | Above the 5 commercial parsers compared | 2026-09-23 |
| OmniDocBench v1.5 | 1,651 pages · table TEDS | 0.892 | — | 2026-09-19 |
| FinTabNet | 300 financial tables · table TEDS | 0.948 | — | 2026-10-12 |
All scores were measured by LEMONG with each benchmark’s public data and scoring code. They are self-measured. Measurement dates are shown.
Input
- HWP·HWPX
- Korean word-processor files, parsed directly
- Digital and scanned
- Word·PowerPoint
- DOC·DOCX·RTF·ODT, PPT·PPTX·ODP
- Excel·CSV
- XLS·XLSX·ODS·CSV·TSV as tables
- Images
- PNG·JPG·TIFF·BMP·WEBP
- Text
- TXT·MD·HTML·XML·JSON
Output
- Markdown
- Headings, tables and figure positions preserved
- JSON
- Element type, page, coordinates and reading order
- TXT
- Plain body text
- REST API
- Upload and retrieve results programmatically
Your documents stay on your servers
- On-premise and closed networksInstalled as containers on a Linux server with an NVIDIA GPU.
- No external AI calls by designLayout and recognition models run on your server, configured to work without internet access.
- Access control and auditAdmin and user roles, encrypted document storage and processing audit logs.
Talk to us
Tell us your document types and volume and your server environment (closed network, GPU). We reply in Korean or English.
support@lemong.ai · 02-6958-6868 · LEMONG Co., Ltd.