Legal Document Conversion: Matching the Format to the Task

Legal document conversion means turning a document from the form it arrived in, often a scanned PDF, into a form someone can work with: a searchable PDF, a Word file, a spreadsheet or structured data. Choose the format by asking who receives the converted copy and what they will do with it.

Picking the format by habit is an avoidable source of trouble. Everything becomes Word, or everything stays a PDF, and then the recipient cannot do the one thing they needed to do. A minute spent on the question "for whom, and to do what?" saves the second conversion.

What does legal document conversion cover?

Legal document conversion covers any change of a document's file format made so that it can be searched, edited, analysed, filed or loaded into another system. It includes scans that need OCR and files that already hold text. The original is not replaced. Conversion makes a working copy beside it.

It helps to sort the work into four kinds, because they call for different tools and different amounts of checking.

Two things are outside the term as used here. Translating a document into another language is a different job with its own risks. And redrafting, where a lawyer rewrites the substance, begins only after conversion has given them something to edit.

A phrase like "legal PDF converter" suggests a single machine that takes any file and returns the right one. In practice the reading step and the output step are separate decisions. Reading happens once and determines how accurate the text is. Output can happen as many times as you need, and determines how useful that text is to a particular person.

One principle runs through all of it: the source file is the record, and every conversion is derived from it. A Word copy of an executed lease is not the lease. A spreadsheet of a bank statement is not the statement. Keep the source, label the copies, and nothing downstream is confused about which is which.

Choosing a format by who receives the converted file

Pick the format the recipient can open and use for their task without converting it again. A court wants a PDF that meets its filing rules. A client usually wants something to read or mark up. An expert wants data. Opposing counsel gets what the rules or your agreement require. A vendor gets what its system loads.

The court

Filing requirements are set court by court. Whether a PDF must be text-searchable, how large it may be, and how bookmarks and exhibits are handled all depend on where you are filing. Read the local rules and the e-filing manual for the court in front of you before choosing settings. The post on searchable PDFs for court filing goes through what courts tend to ask for and how to check a file against it.

The client

Ask what the client will do. To read a set of received documents, a searchable PDF is right, because it looks like the paper and can be searched. To comment on a draft, send Word. Be careful about sending a Word conversion of a signed instrument. A client may reasonably take it for the agreement itself. If you send one, say in the file name and the covering email that it is a working copy made from the scan.

The expert

A forensic accountant asked to work from scanned bank statements wants numbers in cells, not a PDF to retype. Send the spreadsheet and the source PDF together, and say plainly that the figures came from OCR and how far you have checked them. An expert who knows the provenance can decide how much verification their opinion needs.

Not every list is a table. For witness lists, indexes and forms that need to become rows, the post on OCR into Excel covers how to build the columns.

Opposing counsel

In United States federal civil cases, Federal Rule of Civil Procedure 34 lets a request "specify the form or forms in which electronically stored information is to be produced". State courts and other jurisdictions have their own rules, and an ESI order or agreement in your case may say something more specific, so read those before converting anything for production. The post on making a document production searchable covers whether you can require a searchable production and what the rule says when no form is named.

The vendor

A review platform or a printing vendor will tell you what it accepts. Ask for the specification in writing and match it. Schema Law does not produce load files or custom export layouts on any plan. Its eight export formats are fixed, so a load file has to be built in the tool the vendor names, or by the vendor. The post on the Concordance load file explains what such a file contains.

When you do not know the recipient yet

Convert to a searchable PDF first. It preserves the look of every page, adds search, and can be sent to almost anyone. Decide on Word or a spreadsheet when a task calls for one.

Does a document that started in Word or email need OCR at all?

No. A document that was typed on a computer and never printed already contains its text, and OCR can only make that text worse. Convert born-digital files directly, by saving or exporting from the program that made them. OCR is for pages that exist only as pictures.

A file that began as a Word document, an email, a spreadsheet or a PDF saved from any of those stores real characters. The post on the difference between a native PDF and a scanned PDF compares the two kinds of file in detail. Converting it is a matter of Save As or Export. The characters carry over exactly, with no recognition step and no recognition errors.

The mistake to avoid is the paper detour. Someone prints an email, scans the printout, and files the scan. The text that was perfect is now a picture, and getting it back takes OCR and proofreading. Email attachments, client-supplied Word files and PDFs downloaded from a court's system should all be kept in the form they arrived in.

To sort a mixed folder, try to select a sentence in each file and paste it into a plain text editor. If it reads correctly, the text is sound and no OCR is needed. Check a later page as well, since a file can be part typed and part scanned. The post on why you cannot copy text from a PDF explains how to tell whether a PDF has real text.

There are three exceptions where a born-digital original ends up needing OCR anyway. The first is when you only hold a printout or a scan of it, and the native file cannot be obtained. Ask for the native file before you settle for that. The second is a PDF whose text layer is broken, covered in the post on how to convert a PDF to readable text. The third is a signed or annotated version: the Word draft has clean text, but the executed copy with signatures and handwritten changes exists only as a scan, and it is the executed copy that matters.

Converting one scan into several formats for one matter

Convert a scan once, verify the reading once, and export as many formats as the matter needs from that single result. Running the same scan through different tools for different formats gives you several readings that may disagree, and each would need checking separately.

Take a fictional matter. Brightwater Dental is in a dispute with its equipment lessor, and the client sends a 46-page scan: the lease, two amendments and a payment schedule. Over six weeks the file needs four things.

The Word file of that lease is what you draft from. Read each covenant against the scan, including any condition precedent and any waiver printed in the same clause. When the letter discusses a breach of contract, quote the section from the searchable PDF and keep the Excel payment schedule for the figures only.

All four come from one reading of the 46 pages. Schema Law's homepage describes this as one upload and as many exports as you need, with the option to export the same document again later in a different format. The list of export formats shows the eight available.

A sensible order of work:

1. Upload the scan as it arrived and wait for the pages to be read.
2. Review the parsed pages against the scan. Concentrate on party names, dates, amounts and section numbers. This is the one verification that every later export inherits.
3. Export the searchable PDF first and file it next to the source. It is the copy most people will open.
4. Export other formats when a task needs them, not in advance. Unused copies are clutter that someone may later mistake for the working version.
5. Record what was exported, with dates, in a short note in the matter folder.

Formats differ in what they keep. A Word file carries headings and lists. Plain text carries words only. A spreadsheet carries tables. The post on whether OCR keeps formatting describes how each export format treats structure, which is worth reading before you promise anyone a particular layout.

Remember that corrections do not flow between exports. If you fix a misread figure in the Word file, the spreadsheet you exported earlier still has the wrong one. Keep a short list of corrections in the matter note and apply it to each copy you rely on.

Size can get in the way before any of this starts. If the source is too large to convert in one piece, the post on how to convert a large scanned PDF to Word online covers shrinking and splitting it.

Naming converted copies so the source scan stays identifiable

Give every converted copy the source file's name plus a suffix that says what was done to it, and never rename the source. Anyone looking at the folder should be able to tell, from names alone, which file is the original and which copies came from it.

A pattern that works:

<source name> (<what it is>, <date>).<extension>

Applied to a fictional 46-page lease scan that a client called Brightwater Dental sent in:

Rules behind the pattern

• The source keeps the name it arrived with. If your system requires renaming on intake, do it once, at intake, and never again.
• Same stem, different suffix. Sorted alphabetically, the copies sit directly beneath their source.
• Say "OCR" or "converted" in every derived name. That one word tells a reader the text was recognized, not typed by the author.
• Dates in year-month-day order, so that names sort chronologically.
• Page ranges where a copy covers part of the source.
• Status words only when they are true. "Verified" should mean a named person compared it with the scan on that date.
• Two PDFs need the most care. The scan and its searchable copy have the same extension and look identical on screen. The suffix is the only visible difference.

Bates-numbered productions already have identifiers, so keep them. A converted copy of PELL000112.pdf becomes PELL000112 (OCR).docx. Do not invent a descriptive name that breaks the link to the production number.

If your document management system tracks versions, resist saving a conversion as a new version of the source. A version implies the same document, revised. A conversion is a different file type with a different purpose. Store it as a related document, or with the suffix, so that "latest version" never returns a Word reconstruction when someone asked for the executed lease.

When should a converted legal document be converted again?

Convert again when a better source turns up, when the first conversion fails a check, or when the document is about to be used for something that demands more than the first conversion was checked for. Do not convert again merely to get a different format. Export the format from the conversion you already have.

Good reasons to redo the work:

• A better copy arrives. The first scan was a fax of a photocopy, and the client has now found the original, or the other side has produced a cleaner one. Recognition quality follows image quality, so a better image is worth a fresh conversion.
• The native file arrives. If the author's Word file turns up, stop using the OCR copy for drafting. For the wording, the native file is better than any conversion. Check it against the executed scan, since a draft and a signed copy can differ.
• Checks failed. Spot checks found wrong figures or missing pages. Find the cause before repeating: a skewed page, a file split in the wrong place, a page fed upside down.
• The stakes went up. A conversion made so that someone could skim a contract is now the basis for a quotation in a brief or a damages figure. The file may be fine, but the verification was for a lighter purpose. Re-verify at least, and re-convert if doubts remain.
• The document changed. An amendment was signed, or a corrected page was substituted. Convert the new pages and say so in the name.
• Part of the file was never converted. You extracted twelve pages for an early task and now need the rest.

Poor reasons:

• You need Word and you have a searchable PDF. Export Word from the same upload if it is still in your account. If you have deleted the document, upload the source again, never the searchable copy.
• You cannot find the converted copy. That is a filing problem. Fix the naming and it will not recur.
• Someone prefers a different tool. Two tools give two readings. Unless the first reading was poor, a second one doubles the checking for no gain.

When you do convert again, always start from the best source file, not from an earlier conversion. Each pass through print, scan or recognition can only lose detail. Then retire the old copies openly: move them to a "superseded" subfolder or add that word to their names, and update the matter note. The danger in re-conversion is not the new file. It is the old one, still sitting in the folder with a plausible name, and opened by someone in a hurry three months later.