Transcribe Handwriting to Text: Rules for a Transcript

To transcribe handwriting to text that others can rely on, type the page line for line as written, mark every word you cannot read and every alteration in brackets, and have a second person check it against the image. A recognition draft can save typing. The rules and the checking are what make it a transcript.

When is a typed transcript of handwriting worth preparing?

A typed transcript of handwriting is worth preparing when more than one person must read the document closely, when its words will be quoted, or when the handwriting is slow enough to read that each reader would otherwise decipher it again. For a page you only need to find by keyword, a recognition draft is enough.

A transcript is a different product from converted text. Converted text is whatever the software and a quick correction pass produced. A transcript follows stated rules, shows its doubts, names who prepared and checked it, and points back to a specific image. That costs time, so spend it where it pays.

Volume changes the answer too. To transcribe handwritten notes to text across a whole box is rarely justified. Transcribe the pages that carry the dispute and leave the rest as searchable images, with a note saying which is which.

Whether a typed transcript may be filed or shown alongside a handwritten original, and with what label, depends on the court and the kind of proceeding. Check the rules and any standing order that apply to your matter before you attach one to anything.

Using a recognition draft as the starting point for a transcript

A recognition draft is a fair starting point for a transcript when it gets most words right, and a poor one when it does not. Its use is to save keystrokes. It contributes nothing to reliability, because every word still has to be read against the handwriting by a person.

Run the page through a handwriting-capable tool, export Word or plain text, and look at the first ten lines beside the image. The post on converting handwriting to text gives a rule of thumb for when to correct a draft and when to retype. If you are still choosing software, the guide to testing a handwriting to text converter shows how to compare tools on your own pages.

A draft brings one risk that typing from scratch does not. Once you have seen a typed word, you tend to see it in the handwriting. If the draft says "March" and the pen strokes could be "March" or "May 1", your eye settles on "March". Three habits counter that.

1. Read the handwriting first, then the draft. For each line, look at the image, say the line to yourself, and only then look at the typed version.
2. Cover the draft for the hard words. When a word matters, such as a name, a date, an amount or a "not", hide the typed line and write down your own reading before you compare.
3. Reset the line breaks before you correct. Make each typed line end where the handwritten line ends. Software often reflows text into paragraphs, and you cannot check line 12 against line 12 until the two match.

There is a second way to use a draft that suits short, important documents. Type the page yourself from the image without looking at the draft. Then compare your typing with the software's output, using the compare feature in your word processor. Every difference is a place where one of you is wrong. The draft has become a cheap extra reader. It is no substitute for a second human reader, since software and people make different mistakes, but it catches slips of the fingers.

When the sheet is a handwritten last will and testament, you linger on each beneficiary's name, the executor named in the writing, and any codicil crowded into a margin. You read the date and the signature a second time, and you note whether one hand appears to run through the page, the point someone later raises about testamentary capacity. You show the strokes. Counsel decides what they mean.

On Schema Law you upload the scan as a PDF, PNG or JPEG and export the parsed pages. Every plan, the free one included, exports Word, searchable PDF and plain text, and the formats section of the homepage lists the other five. The site says recognition is weaker on handwriting and that nothing marks a passage as uncertain. So the draft will not tell you where its doubts are. Your brackets have to.

How should a transcript mark words nobody can read?

A transcript should mark an unreadable word with a bracketed note in its exact position, such as [illegible], and should never fill the gap with a guess. Use a second, different mark for a word you can read but are unsure of. Put a key to your marks at the top of the transcript.

Square brackets are the usual signal that words come from the transcriber and not the writer. The United States National Archives gives its volunteer transcribers the same instruction in its transcription tips for the National Archives Catalog: "If you can't make out a word, use [illegible]." The same page tells them to put any notation of their own in brackets.

No single standard governs legal transcripts, so choose a set of marks, write them down and use them the same way on every page. One workable set:

A few points of judgment sit behind the table.

Mark doubt at the level of the word. Do not bracket a whole sentence because one word in it is hard. The reader needs to know exactly which strokes are in question.

Give a length for long gaps. "[illegible, about 2 lines]" tells a reader how much is missing. A bare "[illegible]" could hide a word or a paragraph.

Do not expand or correct silently. If the writer put "recieved" or "w/o", so does the transcript. The National Archives page says the same of its own work: "Type words exactly as they are written in the document." If a misspelling might be taken for your typing error, add "[as written]" after it, sparingly.

Numbers get no benefit of the doubt. A figure that could be 1 or 7 is "[1 / 7]", even when the arithmetic elsewhere on the page suggests which. Put the inference in a covering note. The transcript records the marks.

Keep doubt out of formatting. Do not rely on italics, color or highlighting to show uncertain words. They vanish when someone pastes the passage into an email or saves the file as plain text. Brackets survive.

Strike-outs, insertions and handwritten annotations in a transcript

A transcript records strike-outs, insertions and annotations as bracketed notes at the place where they occur, and keeps the struck words when they can still be read. What a writer crossed out or squeezed in afterwards is often the most informative ink on the page. A clean transcript that drops it misstates the document.

The National Archives uses a light version of this in its tips for volunteer transcribers: they "may include [crossed out] next to crossed out text" and use a caret to show added handwriting. Legal work usually needs more detail, because where and how a change was made can matter.

Describe what you see and stop there. "[Second hand]" reports that the letter shapes differ. "[Added later by the supervisor]" is a conclusion about who and when, and it belongs in a memo, if anywhere.

Avoid the word processor's strikethrough formatting for struck words, for the same reason as italics: it is lost in plain text and in many copy and paste operations.

Number the lines, and break each typed line where the handwritten line breaks. Line numbers let a reader go from "statement, line 14" to the image in seconds, and they let two people discuss a reading without confusion.

A short fictional example shows the marks working together.

TYPED TRANSCRIPT. PREPARED AS A READING AID.
Source: Delgado statement, scan p. 2 of 3. Typed by A.N.; checked by P.W.
Key: [struck: ] crossed out; [inserted: ] added; [x?] uncertain; [x / y] either.

11  I got to the loading dock about [6:40 / 6:10] and the
12  gate was [struck: open] [inserted above line: unlocked]. Mr. [Pruitt?]
13  told me the lift had been [illegible] since the
14  weekend. I did not sign the [struck, illegible, 1 word] sheet.
    [left margin, beside lines 13 to 14, in pencil: told T. same day]

Annotations on a typed document are a related but separate job. When the handwriting is a set of comments on a printed contract or pleading, the question of where each note belongs comes first, and the post on handwritten notes covers it under margin notes on a typed page. Turning handwritten annotations to text in that setting usually means a short list of annotations keyed to page and clause, not a line-for-line transcript. For handwritten edits to a signed agreement, see the post on handwritten changes to a contract.

Handwritten affidavits, declarations and pro se filings

A transcript of a handwritten affidavit, declaration or self-represented party's filing is a reading aid, and the handwritten paper remains the sworn or filed document. Transcribe the formal parts with the most care: the date, the place, the statement about perjury or oath, the signature block and anything a notary or clerk added.

Affidavits and declarations

The formal words at the end of a sworn statement are short, often cramped, and easy to treat as boilerplate. They are the part to slow down for. In federal matters, for example, 28 U.S.C. § 1746 on unsworn declarations under penalty of perjury allows a written declaration to stand in for a sworn one when it is subscribed by the person as true under penalty of perjury and dated, in substantially the form the statute sets out. State courts have their own rules on affidavits and declarations, and they differ. Whether a handwritten closing meets the rule that applies is a legal question for the lawyer on the matter. The transcriber's job is to show exactly what the closing says.

That means:

• type the perjury or oath language as written, including errors and omissions, and never "complete" it from the usual form
• show the date as written, with a bracketed alternative if a digit is doubtful
• render a signature as [signature], and add the name only where it is printed or legibly written beneath
• describe a notary's block in brackets, such as [notary stamp, partly legible: commission expires 20[illegible]], and transcribe any handwriting the notary added
• note initials beside each alteration, since they bear on who approved the change

Filings by self-represented parties

Handwritten complaints, motions and letters from people without lawyers, including people in custody, are filed as they were written. A typed working copy helps your team respond. Three cautions apply.

Cite the filed document. Page and line references in your response should point to the document as it appears on the docket. Your transcript's line numbers are for internal use unless the court has been given the transcript.

Do not improve the filing. Keep the writer's wording, spelling and paragraphing. A transcript that turns a rambling sentence into a clean allegation has changed what the party said.

Transcribe everything on the page. Notes on the back, writing around the edge and attachments on scrap paper are part of what was filed. Mark where each sits.

Holographic wills raise further questions about what a typed copy is and is not. The post on handwritten wills covers preparing a working transcript of a holographic will.

Who should check a handwriting transcript, and how?

A handwriting transcript should be checked by a second person who did not type it, working from the image and not from the first reader's memory. The checker reads every word once and gives the doubtful words an independent reading before seeing what the transcriber chose.

Choose the checker for familiarity where you can. Someone who knows the writer's hand, the names in the matter or the trade vocabulary will resolve words a stranger cannot. If the writer is available and it is proper to ask, as with your own client, the writer's reading of a doubtful word is worth recording, marked as such: [Hartley, as read by the writer on 2 May].

A checking routine that holds up:

1. Blind pass on the brackets. Give the checker the image and a list of line numbers for every [?], [x / y] and [illegible], without the transcriber's readings. The checker writes down a reading for each.
2. Compare. Where both readers agree, the question mark can come off. Where they differ, both readings stay in the transcript as alternatives.
3. Full read against the image. The checker then reads the whole transcript beside the handwriting, line by line. For a short, important document, do it aloud in a pair: one person reads the handwriting, the other follows the typed text and stops at any mismatch.
4. Separate pass for names, dates and figures. Go through the page once more looking only at these, character by character.
5. Check the apparatus. Confirm that the header names the right image file, the key matches the marks used and the line numbers run without gaps.
6. Sign and date. Add the checker's initials and the date to the header. Keep the list from step 1 with the file.

The checker's most useful finding is often an unmarked error: a word the transcriber typed with full confidence that is wrong. Those appear in step 3 and nowhere else, which is why the full read cannot be skipped in favor of checking the brackets alone.

How much checking a transcript needs depends on its use. A transcript that will be quoted or given to someone outside the firm gets all six steps. An internal working copy of a long diary might get steps 1 and 2 throughout and a full read of only the pages that matter. Say which in the header, so a later reader knows what "checked" means for the file in front of them.

Store the transcript with the scan it was made from, under matching names. If you used an online tool for the first draft, export what you need and save it to the matter file. The general method for proofreading OCR output against the original is a helpful companion for the typed portions of a mixed page.