OCR Contract Exhibits: Drawings, Forms and Attachments

To OCR contract exhibits, sort them by kind before you convert anything. Typed exhibits go through with the agreement. Tables go to a spreadsheet. Drawings stay as images with a short written description. Attached forms get a marker showing where each begins and ends. Then check that every exhibit the agreement mentions is actually in the scan.

Which contract exhibits does OCR handle badly?

OCR handles exhibits badly when they are mostly graphic, mostly tabular, shrunk to fit the page, or copied several times before they reached the contract. The body of an agreement comes from one word-processing file. Its exhibits come from wherever the parties found them, and their quality varies page by page.

Two causes sit behind most of the trouble. The first is generation loss. A site plan was drawn at full size, reduced on a copier, attached to a draft, printed, signed and scanned. Each pass softened the lettering. The second is orientation. Wide exhibits are often bound sideways, and a page whose text runs up the sheet may come back as noise unless the tool turns it first.

What you can say about any particular tool is limited to what you have seen it do. Run two or three of your worst exhibit pages through it before you commit a whole contract set.

An exhibit that is a form of assignment of rights needs its own pass. Copy the names on its signature lines and the date printed there. If the form includes a warranty, read that paragraph on the image and note any dollar figure. Compare those names with the parties in the preamble before you file the text.

Site plans, drawings and other exhibits with little text

A drawing exhibit gives OCR very little to read, and what it reads comes back out of order. Expect a loose handful of labels, dimensions and title-block entries. Treat the page image as the exhibit, and use the recognized text only as a way to find the page later.

What is worth capturing from a drawing

The title block, usually in the lower right corner, holds the facts that identify the drawing. Read them from the image and type them into your exhibit index yourself:

• drawing or sheet number
• title
• revision letter or number, and revision date
• the firm that prepared it
• the stated scale

These matter where the contract identifies a drawing by number and revision, because someone may later need to show which revision was attached. The lettering in a title block is small, so do not rely on the recognized version of a revision number.

What a scan takes away from a drawing

Color. An agreement may say that the premises are "outlined in red" or that an easement area is "shown hatched in blue" on the plan. If the contract was scanned in black and white, the red line and the blue hatching are now the same gray as everything else, or gone. Search the body for "outlined", "shaded", "hatched", "cross-hatched" and "shown in" and write down what each plan is supposed to show. If the scan no longer shows it, ask for a color scan of that page from the original.

Scale. A plan marked "1 inch = 20 feet" was true at the size it was drawn. After reduction to letter size, and again through a scanner, the stated scale no longer holds. Never measure from the scan. Rely on written dimensions, and say so in your index.

Fine detail. Dimension strings, lot numbers and room labels are the first things to blur. If one of them matters to the question in front of you, read it from the best image you can get, at high zoom, and have a second person confirm it.

Describing a graphic exhibit in the working text

Converted text of a drawing page is close to useless in a Word file. Replace it with one bracketed line, such as "[Exhibit A: site plan, Sheet C-2, Rev. 3 dated 11 May 2022, one page, see scan page 38]". The reader learns what is there and where to look, and a search for "Exhibit A" or "site plan" still finds the spot.

Forms and sample documents attached as contract exhibits

A form attached as an exhibit is a complete document sitting inside another one. It has its own title, section numbers that start again at 1, and blank signature lines. OCR reads the words well enough. The risk is that a reader of the converted text loses track of which document a sentence belongs to.

Examples include a form of guaranty, a form of assignment and assumption, a form of statement of work, a form of estoppel certificate and a form of closing certificate. Each creates the same four problems.

Duplicate section numbers. A search for "Section 5" now has two or three hits: one in the agreement and one in each attached form. A note that cites "s. 5" without naming the document is ambiguous.

Duplicate clause language. A form of guaranty has its own governing law and notice clauses. If you search the converted file for "governed by" and stop at the first hit, you may record the form's choice of law as the agreement's.

Blank signature blocks. The form is unsigned by design. A reviewer paging through the scan can mistake those empty lines for a party's failure to sign the agreement.

Blanks and brackets. Forms are full of underscores, bracketed alternatives and bullet placeholders. Recognition tends to shorten a long underscore, drop it, or turn it into stray characters, so "dated ________, 20__" can come back as "dated , 20".

Marking the boundaries

Fix all four with markers in the working text. At the first line of every exhibit insert a line such as:

[EXHIBIT C BEGINS: Form of Guaranty. Scan pages 41 to 46. Unsigned form.]

and close with a matching [EXHIBIT C ENDS] line. If you work in a searchable PDF and not in Word, add a bookmark at the first page of each exhibit to do the same job.

Restore blanks by hand where they matter. Typing "[blank]" where the form has an underscore is clearer than trying to rebuild the line.

One more distinction belongs in the file note. A form attached to the agreement is a template for a later document. The guaranty that was actually signed is a separate document with its own scan, and it may differ from the form. Do not treat one as the other. If filled-in government or carrier forms are attached, the post on OCR for scanned forms covers boxes, checkmarks and field labels.

Should you convert a contract's exhibits with the agreement or separately?

Do both, for different purposes. Keep one searchable PDF of the whole executed contract, exhibits included, as the file of record. Then pull out the exhibits that need a different kind of output, and convert those on their own.

The single file matters because the agreement refers to its exhibits, and a copy without them is an incomplete copy of what was signed. Splitting is for the working copies only.

Schedules laid out in rows and columns are their own job. The post on converting a scanned contract schedule to Excel covers how to keep the schedule number and clause reference with each row. The grid problems are the same ones that affect exhibits in a court file, which the post on OCR for exhibits with tables explains.

Practical reasons to split

• File size. Drawings scanned in color at high resolution are large. Schema Law accepts a PDF, PNG or JPEG of up to 20 MB, and a contract with a dozen color plans can exceed that as one file while the plans go through easily as their own upload.
• Page limits. The Free, Solo and Team plans each cap the pages in one document, and the pricing page lists the caps. A 300-page agreement with exhibits may need to go up in two parts on a plan capped at 250.
• Export format. A fee table wants a workbook, and the clauses around it want Word.
• Rework. If one exhibit converts poorly, you can rescan and redo that exhibit alone.

Rules for splitting safely

1. Split a copy. The scan as received stays whole and untouched.
2. Cut at the exhibit cover page or first page, never mid-exhibit.
3. Put the exhibit letter and the original page range in the file name: Supply Agreement - Exhibit B - pp 33-37.pdf.
4. Keep an index that lists every part, so the pieces can be counted back against the whole.

Does every exhibit reference in the agreement have a matching exhibit?

Not always, and you find out by making two lists and reconciling them. One list holds every exhibit the body of the agreement refers to. The other holds every exhibit physically present in the scan. Each entry on one list should have a partner on the other.

1. Search the converted body for "Exhibit", "Schedule", "Annex", "Appendix", "Attachment" and "attached hereto". Record each distinct reference with the section that makes it and the description the agreement gives.
2. Page through the scan from the signature page onward. Record each exhibit's label, title and page range from the image.
3. Check the table of contents, if there is one, for a list of exhibits.
4. Reconcile both ways and write down every mismatch.

A worked example, using a fictional lease between Harrowgate Land Co. and Merriton Cold Storage, Inc.:

Three findings come out of five rows. Exhibit B needs a color scan. Exhibit D is absent. Exhibit E is attached but the body never mentions it, which may mean a clause was cut late in drafting or the exhibit was relettered.

Traps in the cross-check

• Misread labels. "Exhibit B" and "Exhibit 8", or "Exhibit I", "Exhibit 1" and "Exhibit l", are easy for recognition to confuse. Confirm labels from the image.
• Sub-exhibits. "Exhibit B-1" and "Exhibit B-2" may be listed in the body only as "Exhibit B".
• Retitled exhibits. The body says "Exhibit C (Scope of Work)" and the page is headed "Exhibit C: Tenant Improvements". Same letter, different title. Note it.
• Placeholders. A page reading "Exhibit F, to be attached" or "intentionally omitted" is present as a page and absent as content. Record which.
• Exhibits to exhibits. An attached form may have its own Schedule 1. List it under its parent.

Exhibits incorporated by reference that are not in the scan

Some documents a contract relies on were never attached, so no scan will contain them. The agreement incorporates them by pointing at them: a policy, a manual, a set of specifications, a proposal, a web page of terms. Find each one, record exactly how the contract identifies it, and go and get a copy.

First separate three situations, because the remedy differs.

When an exhibit that should be attached is missing, check the scan itself before you write to anyone. Sheets get lost in a document feeder, and the post on missing pages in a scanned contract shows how to test for that.

Finding the pointers

Search the converted text for "incorporated", "by reference", "made a part", "in accordance with", "as set forth in", "as amended from time to time", "available at" and "http". Read each hit. Not every one incorporates a document, but the list of candidates will be short.

Pinning down the version

The hard part is which version. Record for each outside document:

• the name the contract gives it, in quotation marks
• any date, version number or revision the contract states
• whether the contract freezes it ("as in effect on the Effective Date") or lets it move ("as updated from time to time")
• where your copy came from and when you got it

Be most careful with web addresses. A page of online terms may have changed many times since signing, and the page you open today is evidence of today's terms only. Ask the client or the counterparty for the version that applied on the signing date, and note that you asked.

Keep these documents in the contract's folder under names that say what they are, such as Referenced - Supplier Quality Manual rev 4 - obtained from client 2024-05-02.pdf. Do not merge them into the executed PDF, because they were not part of what was signed and scanned.

Whether a document is validly incorporated, and which version binds the parties, is a legal question that depends on the contract's wording and the governing law. The job here is to give the lawyer deciding it a complete list and the right papers.