Scanned PDF to Word With Tables That Stay Tables

A scanned PDF converted to Word with tables in it gives you one of three things for each table: a real Word table, text lined up with tabs or spaces, or a picture. Test which you received by clicking inside it. Tabbed text can be rebuilt into a table with Word's Convert Text to Table command.

The difference matters as soon as you edit. A real table keeps each figure in its cell when you widen a column or add a row. Tabbed text holds its shape only until someone changes a word, and a picture cannot be edited at all. This post takes a table from the converted file to a state you can safely type into.

What becomes of a table when a scanned PDF is converted to Word?

A table on a scanned PDF page becomes one of three things in the Word file: a real Word table with rows and cells, lines of text with the columns held apart by tabs or spaces, or a picture of the table. Which one you get depends on the page and on the converter.

On paper a table is ink: some words, some numbers, perhaps some ruled lines. Nothing on the page says "this is a cell". The converter has to conclude that a region is a grid and work out where the rows and columns fall. The post on getting scanned tables into a workbook explains how OCR finds a table on a scanned page. What concerns you here is the result of that guess once it lands in a DOCX.

The three outcomes compared

A single document can contain all three. A ruled fee schedule on page 6 may come through as a table while a borderless list of notice addresses on page 9 comes through as tabbed lines. Judge each table separately.

Tables are one kind of structure among several

Headings, numbered lists and footnotes are inferred from the page in the same way, and each has its own way of going wrong. The post on whether OCR keeps formatting sets out which formatting survives OCR and which does not.

Schema Law's homepage describes its Word export as editable text with its headings, lists and tables. Whatever tool you use, treat the description as a starting point and look at what your own page produced.

How do you tell a real Word table from text lined up with tabs?

Click inside the table. If Word shows its table tabs on the ribbon and pressing Tab moves you to the next cell, you have a real table. If Tab inserts a gap in the line, the columns are only text. If the whole block selects as one object with handles, it is a picture.

That first click settles it for one table. For a document with a dozen schedules, switch on formatting marks and read the page once from top to bottom.

Reading the formatting marks

Select the Show/Hide button, the paragraph mark on the Home tab. Word then displays the characters it otherwise hides.

Four more checks

1. Drag a column edge. In a real table the pointer changes at a column boundary and the whole column moves. In tabbed text there is no boundary to drag.
2. Select a column. In a real table you can select one column from top to bottom. In text, dragging down selects whole lines.
3. Copy it into a spreadsheet. A real table pastes into separate cells. Space-aligned text pastes into a single column.
4. Change the font size. A real table reflows inside its cells. Space-aligned text loses its alignment at once.

A table that is real but wrong

Passing the test does not mean the table is right. Look for two columns merged into one, a column split in two, a row that wrapped onto a second line and became two rows, and a total that landed one column to the left. Count rows and columns against the scan before trusting anything. A table with ruled borders on paper and one fewer column in Word has lost something.

Turning tab-aligned text from a scan back into a Word table

To turn tab-aligned text from a scan back into a Word table, make sure exactly one tab separates each column on every line, select the lines, and use Insert, Table, Convert Text to Table. Word builds one row per paragraph and one column per separator.

Microsoft's instructions for converting text to a table in Word describe the method: insert separator characters such as commas or tabs where the columns divide, use paragraph marks where a new row begins, select the text, and choose Insert, Table, Convert Text to Table. In the dialog you confirm the number of columns, pick an AutoFit behavior, and choose the separator under "Separate text at".

The command is simple. The preparation is where the work lies, because one stray tab or missing separator on a single line shifts every cell after it.

Preparing the lines

1. Turn on formatting marks so tabs, spaces and paragraph marks are visible.
2. Reduce every gap to a single tab. Use find and replace to turn two tabs into one, and repeat until no doubles are left. Where columns are held apart by runs of spaces, replace runs of three or more spaces with one tab, then inspect each line by eye.
3. Count the tabs on each line. A five-column table needs four tabs on every line. A line with three has an empty cell that was skipped, so add a tab where the blank belongs.
4. Rejoin wrapped rows. A long description that ran onto a second line on paper arrives as two paragraphs. Join them so that one row is one paragraph.
5. Remove strays. A page number, a "continued" note or a running header in the middle of the block will become a row of its own.
6. Convert. Select the lines and run Convert Text to Table with the separator set to tabs. Check that the column count Word proposes is the count you expect. A larger number means a line still holds an extra tab.

A worked example

A scanned equipment lease between Corbin Haulage Ltd and Tessway Plant Hire has a payment schedule with four columns: instalment number, due date, amount and balance. After conversion one line reads, with formatting marks on:

3 → → 1 June 2024 → 4,250.00 → 38,250.00

and the next reads:

4 → 1 September 2024 → 4,250.00 34,000.00

The first line has a doubled tab after the instalment number. The second has only a space between the last two figures. Converted as they stand, line 3 would produce five columns and line 4 would put two amounts in one cell. Fix both lines, confirm that each has three tabs, and the conversion yields a clean four-column table.

After converting

Mark the first row as a header row, apply your firm's table style, and right-align the number columns so the decimal points line up. Then read every figure against the scan. Rebuilding a table moves text between cells, and a misread digit is easier to spot once the columns are straight. The post on why OCR gets numbers wrong lists the misreads to look for.

Wide and landscape tables that overflow the Word page

A wide table overflows a Word page when its columns add up to more than the space between the margins. Put the table on a landscape page of its own, fit the table to the window, and then narrow the columns that hold short entries. Change the type size last.

Two kinds of wide table turn up in a scan. One was printed in landscape and bound sideways, such as a rent roll or a cap table. The other was printed in portrait with small type and narrow margins. Neither fits a converted Word file whose page is portrait with ordinary margins.

Fitting the table, in order

1. Give the table a landscape page. Microsoft's guidance on changing page orientation in Word covers a single part of a document: select the content you want on a landscape page, go to Layout, open the Page Setup dialog box, select Landscape, and choose Selected text in the Apply to box. The pages before and after the table stay portrait.
2. Fit to the window. With the cursor in the table, choose the AutoFit option that fits the table to the window. The table then spans the width between the margins and no further.
3. Rebalance the columns. Narrow the columns of dates, codes and amounts to the width of their longest entry and give the spare width to the description column.
4. Reduce the margins on that section if the table is still tight.
5. Reduce the type size by a point or two as the last resort. Small print in a schedule is harder to proofread against the scan.

Pages that were scanned sideways

A landscape table bound into a portrait document can be scanned with its text running up the page. If the converted table is missing or scrambled and the scan shows the page on its side, rotate that page in your PDF program so the text reads left to right, save, and convert the page again. The post on preparing a PDF for OCR covers rotation and the other housekeeping worth doing first.

When the table will never fit

A table of fifteen columns will not be readable on any Word page. Split it into two tables that share the identifying column, as a printed report would, or accept that it belongs in a spreadsheet and refer to it from the Word document as a separate schedule.

Rejoining a table that the scan split across two Word pages

A table that ran over a page break on paper arrives as two tables, or more, with the page furniture of the scan between them. Delete what sits between the pieces so they merge, remove the repeated header and any carried-forward rows, and set the first row to repeat on each page.

On paper the continuation page carried its own copy of the column headings so a reader could follow it, and perhaps a line such as "Balance brought forward". Those were aids for the printed page. In a single Word table they are false rows.

The repair, step by step

1. Show formatting marks and find the end of the first piece and the start of the second.
2. Delete everything between them: page numbers, running footers and headers, "continued" notes and empty paragraphs. When no paragraph is left between two tables, Word joins them into one.
3. Delete the repeated header row that now sits in the middle of the table.
4. Deal with carried-forward rows. A "brought forward" row repeats a figure that is already in the table. Remove it, after noting the figure so you can check the running total.
5. Check the column count across the join. If the second piece has a different number of columns, the converter read the two pages differently. Fix the odd piece before merging.
6. Check the row that straddled the break. A row split across the two pages arrives as two rows, the second with empty leading cells. Merge the text back into one row.
7. Set the header row to repeat. Select the first row and choose Repeat Header Rows.

How the repeating header behaves

Microsoft's page on repeating a table header on subsequent pages says the selection must include the first row of the table, and that repeated headers are visible only in Print Layout view or when the document is printed. It also notes that Word repeats the header at automatic page breaks and does not repeat it if you insert a manual page break within a table. So if a converted file starts each scanned page with a manual break, remove the breaks inside the table and let Word decide where it divides.

Before you call it joined

Count the rows. Add the rows on each scanned page, leave out the repeated headings and carried-forward lines, and compare with the merged table. If the table has a total, add the column and compare with the printed figure. The same problem in a workbook is treated in the section on tables that span several pages.

Should a contract's table stay in Word or move to Excel?

Keep a contract's table in Word when it is part of the text you are drafting and will be read in place. Move it to Excel when you need to sort it, total it or test its arithmetic. Some tables deserve both: the Word copy for the document, the workbook for the checking.

Using both

Word and Excel exchange tables through the clipboard. Convert the table in Word, confirm it is a real table, copy it into a worksheet, and add a sum under each numeric column. Compare your sums with the printed totals. When they agree, the Word table is trustworthy enough to draft from. When they do not, the disagreement tells you which column to proofread.

On a services agreement, the table you check is often the fee schedule or a list of deliverables. Keep the warranty period and any acceptance condition in the columns the page actually prints. If a row states an option, put that date in its own cell so you can sort by it. Skip a column for a term the schedule does not state.

A converter that exports both formats saves the copy step. Schema Law's site says you can export the same document again in a different format, and describes the Excel export as each table on its own sheet with the rest of the text on another. The export formats section lists them all.

Documents with many tables

A long report with tables scattered through it raises a different question, which is how to find and track them all. The post on how to extract tables from a scanned PDF covers page ranges, one sheet per table and an index of what was taken.

Whichever home you choose, label the converted table as taken from the scan, with the page it came from. A schedule in a draft amendment that cites "Schedule 2, page 31 of the executed copy" can be checked by the next person without a search.