Bill of lading OCR and freight document extraction

OCR is the easy part

Plain OCR turns a scanned bill of lading into a block of text. A coordinator still has to find the consignee, the container numbers and the gross weight in it, and type them into the shipment record. docRefinery returns those as named fields: the B/L number as a value, the containers as a table, each with the page and region it was read from. The reconciliation is the work; the typing is the tax on it.

What the shipment record needs from each document

Bills of lading, packing lists and delivery notes are in the 20-type catalogue. Freight documents outside the catalogue, such as customs declarations or certificates of origin, can have a custom classifier trained at no charge.

Freight fields you can check automatically

Many freight fields have a structure, and structure can be checked. A container number is a three-letter owner code, an equipment category letter, a six-digit serial number and a check digit. HS codes are standardised internationally at six digits, and countries append digits of their own to reach 8- or 10-digit national codes. Workflow rules run on every document and can test these: a container number has the right shape, a code has at least six digits, packing list weights add up to the gross weight on the bill. A failed rule flags the run for a person instead of letting the mismatch reach the shipment record.

One carrier, one bill of lading form

Carriers and forwarders each print their own bill of lading and send thousands of them. Learning one carrier's form covers every shipment it handles, and a forwarder's house bill is the same layout from one booking to the next. A team working with a few dozen carriers and forwarders sees a few dozen layouts, each learned once and remembered after that. When a carrier changes its form, the new one is re-trained the first time it arrives. See vendor layout changes.

Handwritten remarks go to a coordinator

Freight documents carry handwriting at exactly the wrong moments: a remark on a delivery note, a corrected weight, a stamped seal number. Fields like these often score below your threshold and go to the review queue, where the coordinator sees the source page with the value's location highlighted, corrects it, and adds or removes table rows. A field that is not on the document, such as a missing notify party, comes back null instead of a guess.

Built for paperwork that crosses borders

A shipment file travels through several companies and countries, so the strengths that matter differ from a finance desk's.

• 70+ languages and scripts, for carriers, forwarders and consignees who don't print in English.
• Shipment packs in one go. A long PDF of bill of lading, invoice and packing list is segmented into its documents, each recognised for what it is.
• One output format. Every freight document returns the same field and table shape, so a new document type adds no parsing work.
• Your TMS's vocabulary. A Return as alias sets the output key, so B/L number can arrive as mbl_no and gross weight as gross_kg.
• Containers and lines as rows, with page and box citations on the cells.
• Customs forms on request. We train, host and re-train a custom type for them without charging for it.
• A workflow per lane or forwarder, started from one sample document, with versions, replay and compare before it goes live.
• Coordinators' fixes reused. Correct a misread once and it can come back as a suggested fix for that carrier's form, proven on the form's own documents first.

From document to the shipment in your TMS

Your code writes the JSON to the shipment (quickstart), or a coordinator exports from the dashboard. Containers and line items stay as rows either way, and the Excel export gives each table its own sheet, ready to match against the booking. The Chrome extension, in early access and not in the Chrome Web Store yet, can auto-fill a carrier or customs web form from a document. There is no transport management system connector or webhook today; integrations are in design on the roadmap.

Freight work that does not need it

• Your carriers already send shipment data as EDI or electronic bills of lading. That data needs no extraction.
• You need goods classified, meaning the right HS code chosen. docRefinery extracts the codes printed on a document; it does not decide them.
• Most of your shipments come from carriers you use once, so few layouts repeat.
• You need tracking events or customs filing. docRefinery reads documents and stops there.