How to Make Deposition Transcripts Searchable

To make deposition transcripts searchable, first test whether the PDF already has text, because a transcript delivered by the reporter usually does. If it is a scan, run OCR and export a searchable PDF for reading plus a plain text copy for searching. Then confirm that page and line numbers still match the original.

That last check is what separates a transcript from most other scans. Testimony is cited by page and line, so text that finds the right words and loses their position has done only half the work. The sections below cover the conversion, the page and line question, condensed transcripts, and how to search several witnesses at once.

Why are some transcripts not searchable to begin with?

A transcript loses its text when it passes through paper. Reporters create transcripts as text, but the copy in your file may have been printed and scanned, faxed, photocopied into a motion exhibit, or produced by another party as page images. Each of those turns typed testimony into pictures of testimony.

The usual sources of an unsearchable transcript are:

• An old matter. The firm kept the bound paper copy and somebody scanned it when the file was closed.
• An exhibit to a filing. Excerpts were printed, marked up or highlighted, and scanned as part of a motion.
• Testimony from another case. A deposition taken in earlier litigation arrives in discovery as images, sometimes several copies removed from the original.
• A copy from a client or co-counsel. They sent what they had, which was a scan.
• A condensed printout. Four transcript pages were printed on each sheet, and the sheets were scanned.

Before you do anything else, test the file. Try to select one word of testimony, or search for a word you can see on the page. The post on why you can't copy text from a PDF explains how to read the result.

If the deposition is recent, the better fix is often a phone call. Ask the reporting firm what electronic formats it can supply for a transcript you already ordered. A file made by the reporter's own software contains the testimony as it was typed, with no recognition errors, and it will always beat OCR of a scan. Use OCR when that route is closed: the reporter is gone, the case is old, or the copy came from someone else.

Converting a transcript PDF to searchable text

Converting a scanned transcript takes one pass of OCR and two exports. Upload each volume as its own file, export a searchable PDF to read and cite from, and export plain text to search and compare. Keep the original scan untouched beside them.

1. Start from the best copy you can get. A clean full-size scan beats a fax or a condensed copy. If you control the scanning, the post on how to digitize law firm paper files covers the settings.
2. Use one file per volume. Name it so a folder sorts sensibly: Rivera_Ana_2024-03-14_Vol1.pdf. Put the witness's surname first and the date in year-month-day order.
3. Upload and wait for the pages to be read. Schema Law takes PDF, PNG and JPEG uploads, with page limits per document that depend on the plan and are listed on the pricing page. If a long volume with its word index passes your limit, split it at a clean page break and note the split in the file names.
4. Export a searchable PDF. You get your original pages with a hidden text layer. In Schema Law the text for each page is attached to that page in reading order. A search takes you to the correct page, and you then read the line number from the image.
5. Export plain text. One text file per volume, with a page break character between the pages of the PDF.
6. Check a sample. Open the first page of testimony, a page from the middle and the last page. Confirm that the page count matches the scan and that a searched word lands on the page where you can see it.
7. File the three versions together: the scan as received, the searchable PDF and the text file.

Step 6 takes five minutes and catches the failures that matter: a volume where recognition stopped partway through, or pages fed to the scanner out of order years ago.

Do page and line numbers survive conversion?

Page numbers usually survive, because each transcript page stays on its own page of the file. Line numbers are less dependable. They sit in a narrow column down the left margin, and OCR may attach them to their lines, move them into a separate block, or drop some. Always take a page and line cite from the image.

The layout is the reason. Court transcript formats show what the numbers on the page are for. The U.S. Court of Federal Claims transcript format requirements, for example, call for pages "numbered consecutively, with a minimum of 25 lines per page" and say each page must "bear numbers indicating each line of transcription". Those are one court's requirements for its own transcripts. Deposition formats follow the reporter and any rules of your jurisdiction, so look at the transcript in front of you. Most will show the same structure: a page number at the top and a numbered line down the side.

To recognition software, that margin is a column of digits beside a column of text. Here is what can come out, and what to do about each:

Two more traps involve page numbers.

The PDF page is not the transcript page. A volume opens with a caption page, appearances and an index, so transcript page 1 may be the fifth page of the file. Volume 2 often continues the numbering from volume 1. Write the offset in the file name or a note, for example "PDF page = transcript page + 4".

Text exports count pages of the file. The seventh page of text in a plain text export is the seventh page of the PDF, not page 7 of the testimony. Apply the same offset.

Test your converted volume with three cites you already trust, taken from a brief or an outline. Look up each one, such as 45:12 to 46:3, using the text to find the passage and the image to confirm the numbers. If all three land, the volume is fit for searching. The post on how to quote text from a scanned PDF covers pin cites and checking the wording of a quotation.

Search the deposition for the witness's name and a phrase you can see in the testimony. A court reporter's PDF is the usual file you start from. The text layer lets you jump to a page and line when you prepare impeachment for trial. Check the line numbers in the margin. If they shifted, the cite points at the wrong answer.

Condensed transcripts with four pages per sheet

A condensed transcript prints four small transcript pages on each sheet in a two-by-two grid. OCR can read it, with two costs. The type is small, so there are more misread characters. And the software has to work out that the sheet holds four separate blocks, or it will read straight across two of them and interleave their lines.

When that reading-order failure happens, a line from transcript page 41 is followed by a line from page 42, then back to 41. Each word is still in the text, so a single-word search works. A phrase search fails whenever the phrase crosses the end of a line, and a pasted excerpt will be scrambled. How software decides where one column ends and another begins is explained in the post on whether OCR can handle two-column documents.

Three practical consequences follow.

• One page of the file is four pages of testimony. A search hit on PDF page 11 means the passage is somewhere among four transcript pages. Read the small page numbers on the sheet to find which.
• The grid order varies. Some condensed layouts run across the top row and then the bottom. Others run down the left side first. Look at the page numbers on the first sheet before you assume.
• The word index at the back is noise. It is many narrow columns of words and page references. It adds false hits to every search, so leave it out of the copy you search if you can.

The better fix is a full-size transcript. Ask the reporter, or whoever sent the condensed copy, whether one exists. If the condensed version is all there is, convert it, use it to locate testimony, and take every quotation and cite from the image after reading it there.

Searching testimony across several witnesses

Once every volume is a searchable PDF with a text copy, you can search all the testimony at once by keeping the files in one folder and using a tool that searches many files. No review platform is needed for a case with a handful of depositions.

There are three ways to do it, in rising order of effort and usefulness.

1. Search the folder of PDFs. A PDF reader with a search-a-folder feature lists the hits by file and page. This is the quickest check for "did anyone mention the March invoice?"
2. Search the folder of text files. Your operating system's file search, or a text editor's find-in-files command, reads text files quickly and shows the matching line with each hit.
3. Build one workbook. For a document with no tables, Schema Law's CSV export lists each line of recognized text with its page number. Open each volume's CSV, add columns for the witness and the volume, and stack them in a single sheet. Filtering the text column for a term then gives you every hit across every witness, with a page to check.

One caution on the third method. If the software detects a table anywhere in the volume, such as an exhibit index, the CSV holds those tables and not the running text. Open the file and look before you rely on it, and fall back on the text file if the testimony is not there.

Search habits matter as much as tools.

• Search for variants. A name may appear as "Ms. Rivera", "Ana" and "the property manager". An unfamiliar name can also be spelled differently from one volume to the next, so check the spelling used in each.
• Check how numbers and dates were written. "March 3rd", "March 3" and "the 3rd" are different strings.
• Use speaker labels. Searching for an attorney's label, such as "MR. OKAFOR:", finds objections and colloquy. "THE WITNESS:" finds answers given outside the question-and-answer flow.
• Search for exhibits by number. "Exhibit 14" pulls together what every witness said about one document. The post on OCR for trial exhibits covers making the exhibits themselves searchable.

A search result tells you where to read. It does not tell you a topic never came up. Misread words and scrambled lines hide testimony, so for anything that matters, read the pages around the hits and the sections where you expected hits and found none.

Exporting transcript excerpts to Word

Export to Word when you need testimony as text you can paste: an excerpt for a brief, a deposition outline, or a table of designations. The Word export gives you the recognized text with each page of the scan starting on a new page, which keeps excerpts easy to line up against the original.

A workable routine for designations:

1. Export the volume to Word and save it as a working copy. Never edit the scan or the searchable PDF.
2. Find the passage and copy it.
3. Paste it into your designation table and type the page and line range from the image, not from the pasted text.
4. Read the pasted words against the image and correct your copy wherever they differ.
5. Delete stray line numbers from inside quoted sentences.

Three points keep an excerpt honest.

The image decides. If the text says "did not" and the page says "did", your copy is wrong. Fix the copy to match the page. If you think the page is wrong, that is a question about the transcript itself, and editing your working file does not settle it.

Turn off automatic corrections. A word processor will happily repair the witness's grammar, capitalize after a dash, and convert quotation marks. Testimony has to stay as it was transcribed.

Check the required format. Courts differ on how designations and excerpts must be presented, including whether whole pages must be attached and how objections are marked. Read your local rules and the pretrial order before you build the table.

What carries over into a converted Word file, and what does not, is covered in the post on how to convert a scanned PDF to Word. The three steps from upload to export are shown under how it works on the home page.