OCR Carbon Copies and Carbonless Multi-Part Forms

OCR can read carbon copies and the pink or yellow plies of a carbonless form when the scan keeps their pale marks apart from the tinted paper. The printed form converts well. The filled-in entries are the weak part, so scan in color, expect doubled or shifted marks, and ask for the top sheet when one exists.

Can OCR read a carbon copy or a carbonless form copy?

Yes, in part. A multi-part form has two layers of content: the form printed by a press, and the entries a person added by pen or impact printer. The printed layer reads like any other print. The entries on a lower ply are faint, soft and sometimes out of place, and they are what you need.

It helps to sort what is on the ply before you convert it, because each kind of mark behaves differently.

The last row matters for a legal file. Two plies of the same form can differ, because each went to a different person who may have written on it afterwards. Treat every ply as its own document.

For typed carbons from a typewriter, where a sheet of carbon paper sat between thin second sheets, the post on typewritten documents has a section on carbon copies, onionskin and mimeograph sheets. The rest of this post is about business forms that make their own copies.

No general description tells you how a tool will do on your form. Schema Law reads pages into text and tables, and its recognition is strongest on clean print and weaker on handwriting. The free plan covers one document a day of up to three pages, enough to run one ply and compare the parsed page with the scan. Nothing marks a passage as uncertain, so the comparison is yours to make.

Why is the pink or yellow copy fainter than the top sheet?

The top sheet carries the pen's own ink. Every sheet under it carries only a chemical mark made by pressure, and the pressure that reaches a sheet falls with each layer of paper above it. The paper color has nothing to do with the faintness. It only tells you where in the stack the sheet sat.

How carbonless paper makes a mark

The National Institute for Occupational Safety and Health describes the mechanism in section 2.1 of its hazard review of carbonless copy paper. In its three-part example the first sheet is "a coated-back (CB) sheet", the second "a coated-front and -back (CFB) sheet" and the third "a coated-front (CF) sheet". The undersides of the top and second sheets carry a layer of microcapsules filled with a colorless dye solution. The review says the capsules "rupture under pressures encountered in normal handwriting or impact printing", and that the released dye moves to the top surface of the sheet beneath, "where it reacts with the clay or resin coating to form an image".

Three consequences follow for anyone scanning these sheets:

• The mark is only as strong as the pressure. A light hand, a felt-tip pen or a gel pen that glides leaves a full line on the top sheet and little on the copies.
• The mark is a dye stain, not a stroke of ink. Its edges are softer than a pen line, so fine parts of letters and digits close up.
• Anything that presses makes a mark. A fingernail, a clipboard clip, a stapler or the edge of a ruler can break capsules. Lower plies collect lines and smudges that no one wrote.

The same review notes that capsules and the reactive coating "can be coated onto the same paper surface", a product it calls self-contained. A sheet of that kind needs no partner sheet to take a mark, so look it over for pressure lines that are not writing.

Why the color of the image varies

The image on a carbonless ply may be blue, purple or black, depending on the paper. Whatever it started as, it sits on a tinted sheet, and neither the image nor the tint is guaranteed to look as it did on the day the form was written. Judge the ply in your hand, not the one you expect. Hold it at reading distance: if you have to tilt the sheet to make out an entry, the scanner will struggle with it too.

Scanning colored carbonless plies in color or grayscale

Scan colored carbonless plies in color at the scanner's full optical resolution, and keep that file as your master. A black-and-white scan has to turn tinted paper and a pale image into pure black or pure white, and on a pink or goldenrod sheet it gets one of them wrong.

Black and white fails in two opposite ways. If the cut-off between black and white sits high, the tinted paper goes dark and the page fills with speckle that swallows the entries. If it sits low, the paper goes white and the pale entries go with it. The post on faded text explains why faint print vanishes in a black-and-white scan in more depth.

Why color and not only grayscale

Grayscale keeps the tones, which is a large improvement. Color keeps one more thing: the difference in hue between a blue image and a pink sheet. Two marks of the same darkness but different colors look identical in grayscale and distinct in color. On a ply where the entry and the paper are close in tone, hue is the only thing separating them.

A capture procedure for a multi-part form

1. Separate the set. If the plies are still joined at a stub, do not feed them as a block. Whether you may tear a set apart is a question about the original, so ask the responsible lawyer before altering it; a set can be scanned ply by ply while still attached, folded back at the stub.
2. Scan each ply singly on the flatbed. Thin paper can jam or crease in a feeder, and rollers add pressure to a sheet that records pressure.
3. Put a sheet of plain white paper behind the ply. These sheets are thin, and the scanner lid or the next page will show through.
4. Scan in color. Save without heavy compression, since compression smears pale marks first.
5. Scan the back. Terms, instructions and payment details are printed on the reverse of some plies and not others.
6. Name the files by ply. Use the form number and the ply color or legend, such as DT-48213_pink_customer.pdf, so nobody later takes one ply for another.
7. Look before you convert. Open the image and zoom in on the faintest entry. If you cannot read it on screen, software will not read it either.

If the copy reached you already scanned in black and white, you cannot add back what the scan dropped. Ask the sender to rescan the paper in color, or ask for a different ply.

Doubled, smeared and offset impressions on a lower ply

A lower ply can show an entry twice, show it shifted out of its box, or show it under a haze of gray. All three come from how the set was handled while someone wrote on it, and all three move text away from the label it belongs to.

Offset entries and field labels

Offset is the fault that does the most quiet harm, because every character may be right and the result still wrong. On a delivery ticket whose plies slipped by one row, the quantity typed for the first item sits on the line of the second. Converted text will pair them as they appear.

Compare the lower ply with the form's own grid. Find one entry whose place you are sure of, such as a date in a date box, and measure how far it sits from where it should. Apply the same shift when you read every other entry. The post on scanned forms covers keeping each answer beside its field label once the text is out.

A worked example

Suppose your client, the fictional Corbel Tile Supply, produces the yellow ply of delivery ticket 48213 to show that 40 cartons reached a job site. On the yellow ply the quantity column reads, to the eye, "40" with a faint second stroke beside the 4. Converted text gives "140". The signature box holds a scrawl and, under it, a printed name the driver wrote in block capitals that comes out as "R MAR1N".

Work through it in this order:

1. Read the quantity on the image at high zoom. The extra stroke is lighter than the digits and has the same shape as the left side of the 4: a doubled impression, not a third digit.
2. Check it against something else on the ticket. If the ticket has an extended price or a carton total, do the arithmetic for 40 and for 140.
3. Record what the ply shows and what you concluded, separately: "Qty reads 40 with doubled stroke; OCR text 140; extended price matches 40."
4. Correct the working copy to "40" and "R MARIN" only where the image supports it, and leave the raw export untouched beside the scan.
5. Ask for the white top sheet, where the driver's pen wrote directly.

Anything that will be quoted or relied on gets read from the image by a person. The general method for checking OCR output against the original applies here with more force than usual.

You see the same doubled digit on the yellow ply of a letter of credit application. Read the amount, the expiry, and the beneficiary named for the draw from the picture. A syndicated loan form in the same pad fades toward the right, where the commitment figures sit, so you read that column from the ply by eye.

Citations, delivery tickets and receipts that arrive as carbonless copies

Carbonless plies reach a law office as the customer's, driver's or recipient's part of a form someone else kept the top of. The examples here are types to recognize in your own file. Which ply a given issuer hands out is something to read off the form, not assume.

Handwritten citations and notices

A citation or notice written by hand in a pad leaves the recipient a lower ply. The printed parts, such as the statute boxes and the court address, convert cleanly. The handwritten parts carry the facts: date, time, location, section number, the officer's name and number. Those are digits and short codes with no surrounding sentence to help, on the faintest layer of the page. Read each one from the image, and compare with the court's or agency's own record when you get it.

Delivery tickets and bills of lading

These are tables: item, description, quantity, weight, with signatures and exceptions noted by hand at the foot. Check the quantity and weight columns digit by digit, look for handwritten exceptions such as a note of damage or shortage written on one ply after the set was split, and watch for offset rows. If you need the figures in a spreadsheet, every plan exports Excel and CSV, the pricing page sets out how many pages each plan converts, and every exported figure still needs reading against the ply.

Receipts, work orders and estimates

A repair order or handwritten receipt mixes a printed price grid with free-hand notes. The signature and any initials beside an authorization line are on the lower ply only as pressure marks, and a light signature may hardly be there. Its absence on your ply does not show that the top sheet was unsigned.

Forms filled in by an impact printer

Multi-part invoices and statements printed on continuous paper have even pressure, so every ply is more uniform than a handwritten one. Their problems are different: thin dotted characters on the lowest ply, and tractor-feed holes or perforations at the edges. The post on OCR for dot matrix print covers those characters.

What to capture besides the entries

• The form number and revision date printed in a corner.
• The serial number of the set, which ties your ply to the issuer's.
• The legend at the foot naming each ply, for example which color is the customer's copy.
• The reverse side.

Which copy of a multi-part form should you ask for?

Ask for the sheet nearest the top of the set, and best of all the top sheet itself, because it holds real ink. When the top sheet is gone, ask whether the issuer scanned or filmed it. When only lower plies survive, get all of them, since each may carry something the others lack.

Reading the form for clues

The form itself tells you what to ask for. Look for a legend at the foot or along the stub that assigns each ply by color. Look at the serial number to confirm two plies are the same set. Look at the back of your ply as well, since printed terms or instructions there may say who each copy was meant for.

A request checklist

1. Identify the form by its name, form number and serial number.
2. Name the ply you hold by its color and legend, and say that its entries are faint.
3. Ask for the top sheet or a color scan of it. State the settings you want: color, each ply as a separate page, front and back.
4. Ask whether the issuer keyed the form into a system, and if so for the record or printout as well.
5. Ask for every other surviving ply, with a note of who held each one.
6. If the answer is that only your ply exists, ask for that in writing.

How you make the request, whether informally, in discovery or by subpoena, depends on the matter and your court's rules, and is a decision for the lawyer on the file.

When two copies disagree

Keep both. Do not merge them into one "best" version. A difference between plies may be a faint mark on one, or a later alteration to the other, and telling those apart is the point of having both. Note for each ply where it came from, scan it in color, convert it separately, and compare the texts line by line. A plain text export of each makes that comparison easy. Schema Law's commitments on how uploaded documents are kept and deleted are listed under security on the home page, which is worth reading before a client's records leave your office in any form.