OCR Software PDF to Word: How to Choose a Converter

Choose OCR software for PDF to Word work by the Word file it hands you, not by its feature list. The best converter for your office is the one whose output, on your own scans, opens as flowing paragraphs with real headings, lists and tables, and needs the least repair before a lawyer can edit it.

No ranking can tell you that in advance, because the answer depends on the documents. What a buyer can do is fix the criteria, run the same five documents through each candidate, measure the cleanup and then look at price.

Judge the Word file by the edit a lawyer will make next. A contract should keep its section numbers. A pleading should keep its headings. A deposition transcript should keep questions and answers in order. Have an attorney or a law clerk repair one page of each and write down the minutes. That time is the real price, and checking it is part of your duty of competence.

What separates PDF to Word OCR software that gives a usable document from one that does not?

A usable Word document from PDF to Word OCR software has the right characters, in reading order, in paragraphs that reflow when you edit them, with structure built from Word's own features. An unusable one may look identical on screen and fall apart at the first keystroke.

Four tests, in the order they matter:

1. Are the words right? Wrong characters in names, dates, amounts and section numbers make every later quality irrelevant.
2. Is the order right? Text from a two-column page, a footnote or a margin note should not be spliced into the middle of a sentence.
3. Does the text flow? Delete a sentence and watch whether the paragraph closes up and the page break moves.
4. Is the structure real? A heading should be a heading, a numbered clause a list item, a table a table.

The first two are about recognition. The second two are about how the software writes the Word file, and they are where converters that read equally well produce very different documents.

A quick way to see what you were given: open the converted file and switch on Word's formatting marks, the button that looks like a paragraph sign. The marks expose what the page view hides.

Keep one distinction in mind while you shop. Looking like the scan and being editable pull in opposite directions, and the product that produces the more faithful picture of the page is not for that reason the better drafting tool. The step-by-step conversion, from upload to checked file, is set out in the guide to converting a scanned PDF to Word.

Flowing paragraphs or text boxes: two ways PDF to Word OCR software rebuilds a page

PDF to Word OCR software rebuilds a page in one of two ways. It either writes the text as ordinary paragraphs that flow from margin to margin, or it places each block in a positioned box so the Word page mirrors the scan. Flowing output is easier to edit. Boxed output is closer to the original's appearance.

For a buyer the question is which style a product produces, whether you can choose, and whether the choice is per document.

How to tell which one you have. Click in the middle of a paragraph and press Ctrl+A. In a flowing document the whole text selects as one run. In a boxed document only the box you clicked in selects, or the selection skips between islands. Then type three lines into the middle of the first page. In a flowing document everything after it moves down. In a boxed document the new text overflows its box, hides behind the next one or pushes it off the page.

Why it matters for legal editing. Tracked changes, compare, automatic numbering and cross-references all assume one continuous body of text. A redline between a boxed conversion and a later draft can report changes that are only layout. A flowing conversion gives up some resemblance to the scan and behaves like a document someone typed.

When the boxed style is the right one. A form with fixed fields, a certificate, or a page where position carries meaning may be better served by output that holds everything in place. If you will only read and lightly annotate the Word file, resemblance may matter more than editability.

Put three questions to each candidate's documentation or trial:

• Does it offer both styles, or only one?
• Is the style a setting you choose at export, or fixed?
• Does the flowing style keep headings, lists and tables, or flatten everything to plain paragraphs?

Schema Law's homepage describes its Word export as "Editable text with its headings, lists, and tables". That sentence is all the site says about layout, so if you need the Word file to mirror the scan, run a page through the Free plan and look before you assume either way. The choice between the two styles for a particular document, made at conversion time, is covered under flowing text or a page-for-page layout in the Word output.

Which Word features should a PDF to Word converter rebuild for legal drafting?

For legal drafting, a PDF to Word converter should rebuild headings, numbered clauses, tables and footnotes as Word's own features, and keep running headers and page numbers out of the body text. Rank those by how much of your work depends on each, because no converter is certain to rebuild all of them on every scan.

Three points for a legal buyer.

Numbering is the hard one. A scan shows "4.2(b)" as ink. Whether the converter writes that as a list item three levels deep or as typed characters decides how much work a redraft will be. If your office restyles every incoming document with its own numbering template anyway, typed numbers in clean paragraphs may be all you need. Decide which camp you are in before you score this row.

Tables deserve their own trial. A table can arrive as a real Word table, as tab-aligned text or as a picture. The differences, and what to do about each, are the subject of the post on scanned PDF to Word with tables.

Things that are not text stay pictures. Signatures, seals and letterhead logos cannot be rebuilt as Word features. What you can ask is whether the converter keeps them as images in place, drops them, or tries to read them as words.

Mark each row as required, useful or irrelevant for your practice. A litigator converting opposing briefs to quote from needs footnotes and little else. A transactional lawyer redrafting leases needs numbering and tables. Their best converters may differ.

A five-document Word conversion test for a legal buyer

A five-document Word conversion test runs the same five scans, chosen to stress different Word features, through every candidate and scores each output with the same four checks. Five documents cover the Word features legal drafting depends on, and the set is small enough to rerun for every candidate.

Choose the five from your own closed files. Use documents with no restriction on outside handling, or redact a copy first.

Then, for each candidate:

1. Convert all five with default settings. Note any setting you had to change, because everyone who uses the product afterwards will have to change it too.
2. Learn the output style on document 2. Click in a paragraph and press Ctrl+A to see whether the whole text selects, then type three lines into page one and watch whether the rest moves down.
3. Check the Word features your work requires: headings in the navigation pane on 2, a footnote mark that jumps to its note on 3, Word's table tools on 4.
4. Proofread one full page of each against the scan, marking every wrong character in a name, date, amount or section number.
5. Time the repair of one page of document 2 until you would hand it to a partner.
6. Record the result on one sheet per candidate.

A fictional score sheet shows the shape. Reyes Hartmann LLP tests two candidates. Candidate A produces flowing text, lists the headings in the navigation pane, types the clause numbers as characters, and needs four minutes of repair on the contract page. Candidate B mirrors the page in boxes, looks closer to the scan, and takes eleven minutes to make the same page editable. For a firm that redrafts, A wins despite looking plainer.

Keep the five scans and the score sheets. They are the firm's own benchmark, and rerunning them a year later tells you whether a product has changed. This test looks only at the Word file. A wider trial covering other formats and volume is described in the post on how to test OCR software.

How much Word cleanup per page is acceptable from OCR software?

Acceptable Word cleanup from OCR software is whatever stays well under the time it would take to retype the page, by a margin you set in advance. Pick the number before testing. Without a threshold fixed beforehand, every result looks tolerable once you have paid for the product.

Work it out from your own office.

1. Time a retype. Have the person who would do the work type one dense contract page from the scan, then proofread it. That is the ceiling.
2. Set the threshold as a fraction of the retype time, a quarter for example. The figure is yours to choose. Write it down.
3. Measure cleanup the same way. Structure repair plus proofreading, on the same page, by the same person.
4. Compare by document type. A product can pass on clean contracts and fail on faxes.

Separate two kinds of cleanup, because they have different costs.

Structure repair is mechanical: removing line-end paragraph marks, restoring numbering, rebuilding a table. It is tedious and visible. A person can see when it is finished.

Verification is reading the text against the scan for wrong characters. It is invisible work, and it does not shrink to zero with any product. Schema Law's homepage says recognition is strongest on clean print and weaker on faint faxes, stamps over text and handwriting, and that nothing marks a passage as uncertain, so names, dates and amounts have to be checked against the original. Whatever a vendor says about its own product, budget verification time for every page that will be relied on.

A sample threshold table to adapt:

If a candidate exceeds your threshold on the documents you convert every week, it has failed, however well it does on the baseline letter.

Paying for PDF to Word OCR software by the document, the page or the month

PDF to Word OCR software is sold per document, per page or by monthly subscription with a page allowance. For a single job, the right model follows your pattern: occasional long documents favor paying per use, a steady weekly flow favors a monthly plan sized to your pages.

Match the model to the pattern first, then compare prices within it.

Four limits decide whether a plan fits Word work in particular.

Pages per document. A 180-page lease has to fit in one upload or be split, and splitting a contract mid-clause makes the Word files harder to stitch together. On Schema Law the limit is 3 pages per document on Free, 250 on Solo and 1,000 on Team.

The monthly allowance. Solo includes 1,000 pages a month at $49 and Team 10,000 at $199. Count the pages you convert, including repeats.

Users. Where a vendor charges by the seat, a team's cost is the plan price times the number of people who convert. Ask how users are counted before you compare monthly prices.

Whether Word is in the plan at all. A plan can hold formats back for a higher tier, so read the format list for the tier you would buy. Schema Law includes every export format, Word among them, on every plan, Free included; the list of formats is under export formats on the homepage, and the plan lines are on the pricing page.

Count conversions that fail your threshold as cost too. A cheap plan whose output needs eleven minutes of repair a page costs more in staff time than the subscription saved. The arithmetic for turning any of these models into one comparable figure is in the post on OCR cost per page.

Two neighboring choices are worth ruling in or out before you buy. If your volume is low enough that paying may be unnecessary, read how to judge the best free OCR PDF to Word converter. If what you really need is the original scan made findable, with no editing, a Word converter is the wrong purchase, and the post on the best searchable PDF software covers that decision.