Skip to content
⁨NDDev AI⁩, home page
Community
Language: English
Display
Theme
Motion

Document recognition and data extraction

NDDev recognises documents and extracts their data for accounting, logistics and procurement teams: invoices, delivery notes, contracts, scans and photos become fields for your accounting system, and every value keeps a link to its place in the source file. Declaro, a product we build as a technology partner, works this way to prepare draft customs declarations.

Who it is for

Accounting, logistics and procurement teams, customs brokers, banks and insurers who copy data from incoming documents into business systems by hand.

  • Scans, PDFs and office files
  • A source link for every field
  • Rule-based checks
  • Export to your system

What happens to a document

We detect the file type from its content, because extensions often mislead. Text PDFs and Word and Excel files are parsed directly; scans and photos go to recognition, with the engine chosen by benchmark on your documents. A failure in one file stays in that file, and the rest of the batch carries on.

  • parsing of PDFs, office files, scans and photos
  • a recognition engine chosen by benchmark
  • field extraction with a quote from the source
  • checks on totals, dates and required fields
  • a review screen for your staff
  • export to JSON, XML or your accounting system
  • per-field accuracy measured on a reference set

The model turns each document into a record with typed fields and the quotes from the text each value rests on. Code then assembles the result and checks it against rules: totals add up, dates fall within range, required fields are filled. The result goes to your system as JSON, XML in your schema or a file for your accounting software.

In Declaro this pipeline assembles a draft customs declaration and XML validated against an XSD schema. Every field leads back to its source, and the declarant signs: the system prepares the draft, a person checks and signs it.

Rolling out recognition on your documents

We measure accuracy for every field on your documents, and you see those numbers before integration.

  1. Reference set

    We collect real documents and fill in the correct field values by hand to measure accuracy against.

  2. Pipeline

    We set up parsing, recognition, extraction and validation rules for your document types.

  3. Measurement

    We calculate accuracy for every field and show where and why the system makes mistakes.

  4. Integration

    We connect the export to your system and the review screen for your staff.

  5. Ongoing additions

    New document types and errors we find go into the reference set, and accuracy is recalculated after every change.

Work in this area

  • Platform

    Declaro: software for customs declarants

    Declaro prepares a draft customs declaration from shipment documents, with a link to the source of every value: the declarant sees which document each value came from, checks the draft and signs the declaration themselves. Declaro is in development. It is a product of a separate company where NDDev's CEO, Danil Silantyev, is the CTO.

Questions about document recognition

Can the system cope with poor scans?

Recognition quality depends on the scan, so we test it on your documents before work starts and show the result for every field. Fields the system is unsure about go to a staff member for review.

How is this different from ordinary OCR?

OCR returns text. We extract specific fields from it, link each one to its source, check them against rules and export them in the format you need. Your staff receive a ready record to review.

Which documents have you worked with?

Declaro prepares a draft customs declaration from documents, plus XML for the customs system. We also have a project for a public-sector customer in customs, with details covered by an NDA.

Can documents be processed on our servers?

Yes. We deploy recognition and the language model on your hardware, and the documents stay with you. If an external provider is acceptable, we agree in advance which fields it may receive.

What does the price depend on?

The minimum sum is shown in the price block on this page. The final price is set by the number of document types and fields, scan quality, validation rules and integrations. We fix the price and timeline in the specification.

from $6,700

Free warranty
60 days

Prices are starting prices for information only and are not a public offer. The final price is fixed in the specification.

Documents, video and forecasting

Send us a few typical documents

Anonymised samples and a list of the fields you need help us estimate accuracy and give you a price.

We reply within 24 hours

Assistant

Loading the assistant…