DEEP OCR

Any page in, every value out.

Reads documents with handwriting, Hanja, blurry photos, and fax noise, including their structure and context, and delivers data your business systems can use right away.

Contact usStart for free

91%

Claims with no human touch

At one insurer, the share of claims closed without human handling rose from 28% to 91%.

780

Pages per minute

Reads more than 780 pages per minute with multithreaded parallel processing. Speed stays steady even when documents arrive all at once.

400M+

Training pages

Trained on more than 400 million text images, so it starts on customer documents without separate labeling.

80+

Sites in production

Reads real operational documents at more than 80 sites, including finance and the public sector.

What it reads

Trained on the documents people actually send.

Trained on real field documents. It reads text as written, including handwritten notes, old typewritten documents, Hanja newspapers, formulas, dark-printed tables, and fax noise.

Handwriting document example

Handwriting

Handwriting

Typewritten document example

Typewriter

Typewritten

Hanja document example

Hanja

Hanja

Math and symbols document example

Math & symbols

Math and symbols

Printed document example

Print

Printed

English script document example

English script

English script

Low light and blurred documents document example

Low light & blur

Low light and blurred documents

Fax noise document example

Fax noise

Fax noise

These are real documents that KDL has published as DEEP OCR reading examples.

How it reads

Understands the page, then checks every value against it.

A document-specific VLM designed by KDL reads the layout, tables, and the relationship between keys and values, along with the text. Each value read is checked again against the original, and only values with low confidence go to a person.

  1. 01

    Understand

    Reads the structure and context of the document, not just the characters, and finds values by meaning, such as contracting parties, contract period, and signature fields.

  2. 02

    Verify

    With a two-stage OCR Hybrid structure, OCR cross-checks where each value the VLM found appears in the original. This prevents values that do not exist from being made up.

  3. 03

    Route

    Results with low confidence are filtered out automatically so a person can check them against the original. The rest move to the next step without human handling.

Medical Fee Statement and Receipt

Reading 0 / 169
Medical Fee Statement and Receipt sample document

Field results

Treatment information

OutpatientUnchecked81%
InpatientChecked97%
DischargeChecked80%
InterimUnchecked92%
Patient registration no.71091083%
Patient nameOh Min-seo95%
Year202595%
Month0561%
Day0367%
Year202597%
Month0599%
Day0399%
Night visitUnchecked97%
Holiday visitUnchecked81%
DepartmentSurgery86%
DRG codeEZ901288%
Ward2-bed room99%
Patient typeNational Health Insurance89%
Receipt no.202509-99537797%

Consultation fee

Partial copay, patient share095%
Partial copay, NHIS share082%
Full patient share084%
Non-covered082%

Room fee, private room

Partial copay, patient share085%
Partial copay, NHIS share095%
Full patient share093%
Non-covered094%

Room fee, 2–3 bed room

Partial copay, patient share150,00091%
Partial copay, NHIS share450,00093%
Full patient share094%
Non-covered096%

Room fee, 4+ bed room

Partial copay, patient share087%
Partial copay, NHIS share095%
Full patient share095%
Non-covered090%

Meals

Partial copay, patient share42,00086%
Partial copay, NHIS share42,00084%
Full patient share061%
Non-covered095%

Dispensing fee, service

Partial copay, patient share088%
Partial copay, NHIS share083%
Full patient share080%
Non-covered094%

Dispensing fee, drugs

Partial copay, patient share092%
Partial copay, NHIS share098%
Full patient share085%
Non-covered092%

Injection fee, service

Partial copay, patient share073%
Partial copay, NHIS share094%
Full patient share065%
Non-covered098%

Injection fee, drugs

Partial copay, patient share096%
Partial copay, NHIS share091%
Full patient share085%
Non-covered094%

Anesthesia fee

Partial copay, patient share90,00096%
Partial copay, NHIS share270,00081%
Full patient share090%
Non-covered081%

Procedures and surgery fees

Partial copay, patient share280,00095%
Partial copay, NHIS share840,00093%
Full patient share0⚠ 46%
Non-covered083%

Lab test fees

Partial copay, patient share081%
Partial copay, NHIS share0⚠ 56%
Full patient share085%
Non-covered0⚠ 50%

Imaging fee

Partial copay, patient share084%
Partial copay, NHIS share091%
Full patient share080%
Non-covered090%

Radiation therapy fee

Partial copay, patient share090%
Partial copay, NHIS share098%
Full patient share086%
Non-covered094%

Medical supplies

Partial copay, patient share70,00089%
Partial copay, NHIS share210,00089%
Full patient share085%
Non-covered120,00091%

Rehabilitation and physical therapy fees

Partial copay, patient share090%
Partial copay, NHIS share098%
Full patient share097%
Non-covered080%

Psychotherapy fee

Partial copay, patient share090%
Partial copay, NHIS share092%
Full patient share092%
Non-covered087%

Whole blood and blood components fee

Partial copay, patient share091%
Partial copay, NHIS share098%
Full patient share090%
Non-covered097%

CT diagnosis fee

Partial copay, patient share0⚠ 49%
Partial copay, NHIS share084%
Full patient share087%
Non-covered090%

MRI diagnosis fee

Partial copay, patient share081%
Partial copay, NHIS share083%
Full patient share090%
Non-covered092%

PET diagnosis fee

Partial copay, patient share088%
Partial copay, NHIS share095%
Full patient share060%
Non-covered099%

Ultrasound diagnosis fee

Partial copay, patient share087%
Partial copay, NHIS share089%
Full patient share077%
Non-covered092%

Prosthetics and orthodontics fee

Partial copay, patient share096%
Partial copay, NHIS share096%
Full patient share091%
Non-covered096%

Certificate issuance fee

Partial copay, patient share096%
Partial copay, NHIS share080%
Full patient share093%
Non-covered086%

Selective benefits

Partial copay, patient share092%
Partial copay, NHIS share083%
Full patient share098%
Non-covered085%

Flat rate, age 65+ and others

Partial copay, patient share096%
Partial copay, NHIS share093%
Full patient share088%
Non-covered060%

Flat-rate fee, nursing hospital

Partial copay, patient share090%
Partial copay, NHIS share089%
Full patient share094%
Non-covered086%

Flat-rate fee, palliative care

Partial copay, patient share0⚠ 55%
Partial copay, NHIS share0⚠ 52%
Full patient share070%
Non-covered085%

DRG bundled payment

Partial copay, patient share099%
Partial copay, NHIS share097%
Full patient share085%
Non-covered093%

Other

Partial copay, patient share075%
Partial copay, NHIS share0⚠ 55%
Full patient share081%
Non-covered099%

Total

Partial copay, patient share632,00085%
Partial copay, NHIS share1,812,00084%
Full patient share089%
Non-covered120,00099%
Amount over cap094%

Fee calculation details

Total medical fees2,564,00091%
Total NHIS-covered amount1,812,00096%
Total patient-borne amount752,00092%
Amount already paid084%
Amount payable752,00084%
Card083%
Cash receipt752,00085%
Cash083%
Total752,00091%
Unpaid amount084%
ID verification no.965539361494662%
Approval no.779407663151859298%
IssuanceChecked95%

Type of care institution

Clinic-level health centerUnchecked94%
HospitalUnchecked94%
General hospitalChecked95%
Tertiary general hospitalUnchecked85%

Institution information

RepresentativeJeong Seo-yeon86%
Business registration no.110-82-3344583%
Business nameSeoul Central General Hospital73%
Phone no.02-2200-700098%
Address95 Eulji-ro, Jung-gu, Seoul94%

Issue date

Year202597%
Month0898%
Day1885%
Procedures and surgery feesFull patient share0⚠ 46%
Lab test feesPartial copay, NHIS share0⚠ 56%
Lab test feesNon-covered0⚠ 50%
CT diagnosis feePartial copay, patient share0⚠ 49%
Flat-rate fee, palliative carePartial copay, patient share0⚠ 55%
Flat-rate fee, palliative carePartial copay, NHIS share0⚠ 52%
OtherPartial copay, NHIS share0⚠ 55%

Workflow

From scan to system in three steps.

Turns scanned PDFs and images into searchable, usable data by splitting them into tables, paragraphs, and keys and values. Just define the fields, and you get results in the same format even from documents with different layouts.

Upload.

Upload documents with differing formats as they are, such as invoices, medical fee statements, and handwritten notes.

Run.

Analyzes each document and organizes only the values you need. You can set in advance which items to extract for each document type.

Connect.

Sends the organized data directly to your in-house ERP, Excel, or generative AI through an API.

Document AI Packs

Ready for the paperwork your industry runs on.

Document AI packs tailored to each document type. Without building templates, they extract the key items and connect them to your business systems.

Bills of lading (BL), invoices, certificates of origin reading example

Trade documents

Bills of lading (BL), invoices, certificates of origin

TradeSettlement
Financial loan applications reading example

Loan applications

Financial loan applications

ExtractionUnderwriting
Welfare claim applications reading example

Welfare claims

Welfare claim applications

EligibilityClaims
Retirement pension certificates reading example

Pension certificates

Retirement pension certificates

EvidencePension
Account opening applications reading example

Account opening

Account opening applications

KYCOnboarding
Air waybills reading example

Air waybills

Air waybills

WaybillCustoms
Shareholder registers, term sheets reading example

Shareholder registers

Shareholder registers, term sheets

Bulk recordsAudit
Process manuals reading example

Process manuals

Process manuals

KnowledgeStandards
Medical records, prescriptions reading example

Medical records

Medical records, prescriptions

Sensitive dataInsurance

The blue boxes are areas DEEP OCR found in the document. These are reading examples published by KDL.

Hyundai Capital

752 loan document types, read and sorted on arrival

Automated the classification and reading of 752 types of Hyundai Capital loan documents.

Case · Hyundai Capital

752 kinds of loan paperwork, sorted and read on arrival.

Hyundai Capital sorts and reads 752 types of loan documents in a single system the moment they arrive. As soon as a document comes in, its type is identified and the values needed for review are read out.

752

Loan document types

Hyundai Capital

91%

Claims with no human touch

I Insurance

70%

Less input time

N Financial

See Finance

Security

Runs inside your closed network.

Runs just as well in on-premise and closed-network environments fully cut off from external networks. Documents and models never leave your organization.

Software quality certification (GS) Grade 1

GS certified

Software quality certification (GS) Grade 1

Quality management certification

ISO 9001

Quality management certification

Women-Owned Business Certification

Women-owned business

Women-Owned Business Certification

Innovative Procurement Product by PPS

K-Innovative Product

Innovative Procurement Product by PPS

Choose how it runs

On-Premise

Installed directly on your own servers and operated there.

Fits when

  • Servers are ready
  • External internet cannot be used
  • Security matters most

Private Cloud

Operated in a dedicated cloud with data kept separate.

Fits when

  • You want a closed-network cloud
  • A dedicated environment and data separation are required
  • You want to hand off server operations

DEEP Hub

Install dedicated software as a package, even without an internal server.

Fits when

  • You have no internal servers
  • Buying servers all at once is a burden
  • You want to start using the software right away

SaaS

Start right away in the cloud, with no installation.

Fits when

  • You want to use it right away with no installation
  • You can use the internet
  • You need a reasonable price

Works with the rest of DEEP Agent OS.

DEEP Parser

Understand

Tables and layouts, kept as structure

Explore

DEEP Agent

Act

Agents that process, answer and analyze

Explore

DEEP Ops

Operate · Workflow

Keeps every agent accurate after launch

Explore

Send us your hardest page.

Share a document that is hard to read. We run a PoC with your actual documents and first confirm how much of it can be read.

Contact usStart for free