1 The Basics
1.1 Work offline
1.2 Listen to the audio
2 Editing the Content
2.1 Speaker Labels
2.2 Resolve [?] and [xx] if possible
2.3 Timestamp any remaining inaudibles
2.4 Verify spelling, proper names, and unusual words
3 Preparing the Transcript
3.1 Adjust TC grades
3.2 Remove overlap and splices
3.3 Certify as deliverable
4 Special Situations
4.1 Special instructions from the customer
4.2 Foreign languages
4.3 Incorrect or missing text within a TC
5 Edit Grading
This is a supplement to the Main Style Guide. All the rules in that document apply to edit jobs, unless otherwise specified here.
Just as with transcription jobs, it is a good idea to do most edit assignments offline. You can download the audio file to use with your audio player of choice, and copy the text into a word processor that will allow you to spell check and save as you go.
When you are done, replace the text in the job with your final text to review and use the tools on the job page to make any final tweaks.
Editing is much more than just cutting out splices and timestamping inaudibles. You are expected to listen to the entire audio, correcting errors in the transcripts and carrying out the other steps described below.
Mike's first rule of editing: if it ain't broke, don't fix it.
If you make changes that take something the transcriber got right and make it wrong, your work will be heavily downgraded or rejected, and you will lose badge points.
See the Main Style Guide for the basic rules regarding speaker labels.
In most cases, the only way to fix an edit with speaker label problems is to go back to the original TCs; in other words, we have to discard the edit and start over. Therefore, speaker label problems are cause for rejection, no matter how good the rest of your work is.
Different transcribers may refer to the same speaker in different ways, for example, "Woman 1," "Woman 3," "Interviewer," and "Sandy." You must choose the most appropriate label for each speaker, and make sure it is applied consistently throughout.
Transcribers may use generic labels because they don't have any more information about a speaker, based on their individual chunk of the audio. Listening to the full audio often provides you with enough information to improve the speaker labels.
If you know a person's name, use the name throughout. If you don't know their name, use their role, if known. Use "Man" and "Woman" labels as a last resort.
See the Main Style Guide for more detail on choosing appropriate speaker labels.
Remove numbers from any speaker label that doesn't need them. Transcribers are instructed to number all "Man" and "Woman" labels, even if there is only one speaker with a "Man" or "Woman" label in their particular chunk.
As you improve speaker labels and make them consistent, you may find that there is only one person with a "Man" or "Woman" label in the entire audio. If there is no "Man 2" or "Woman 2," then remove the number.
The same goes for any other numbered speaker labels. If you have a "Interviewer 1" but no "Interviewer 2," then remove the number so that speaker is simply labeled "Interviewer."
Follow the Main Style Guide for details on labeling speakers in Group Style transcripts. If Group Style is approved for a transcript, it must be used throughout the document.
If Group Style is not approved, each speaker must be tracked, and there must be no "Participant" labels of any kind (unless specifically requested in customer notes).
Before submitting, check the speaker list to make sure your work doesn't have any of these problems:
The same name spelled two different ways
The same participant labeled two different ways
More different speaker labels than there are speakers
"Participant" labels in a non-Group Style transcript
Gaps in numbered speaker labels ("Man 1" and "Man 3," without a "Man 2")
Numbered speaker labels where there is a 1 but no 2
Pay special attention to spots marked with [?] and [xx], and resolve them with the correct text if possible. Any [?] or [xx] that you cannot resolve must be changed to timestamped [inaudible] or [indecipherable] tags (see next section).
It is OK to delete words that the transcriber typed and replace them with an [inaudible] tag, if you really believe the transcriber got it wrong, and you can't figure out what was actually said.
When you are done, the transcript should contain no [?] or [xx] tags, whether certified for delivery or not. If it does, your work will likely be rejected, regardless of how good the rest of it is.
If timestamps have been ordered for a particular job, the transcript will go through a separate process to insert them after the final edit is complete. Therefore, editors timestamp only inaudible tags, even if customer comments mention timestamps at other intervals.
Your inaudible tags should be in mm:ss format (not hh:mm:ss), and should not use leading zeros. See examples below. The timestamp should correspond to the first syllable of the inaudible content.
Woman: We thought it was too [inaudible 3:09] for a Friday.
Interviewer: What are your thoughts on [inaudible 89:39]?
Geordi: I need to reverse the polarity of the [inaudible 103:21].
Your audio player may show the time elapsed in hh:mm:ss format instead of the total elapsed minutes. Most players will allow you to set your preferences to show mm:ss instead.
Transcribers are expected to verify as much as possible before submitting but, as the editor, you are responsible for making sure the final product is both correct and consistent. Be especially careful with company names, which often have unusual spelling, capitalization, and punctuation.
Do web searches as needed, and check the websites of any companies or organizations mentioned to verify their preferred spellings, names of employees mentioned in the audio, etc.
Transcribers may have left comments about items that they were (or were not) able to verify. Check for comments by hovering your mouse pointer over the TC numbers in the grade list.
If you are unable to verify spelling but are sure that you have the right word or name, use your best guess at the spelling, remove all [sp] tags, and leave a comment.
If you are not even sure what the word or name should be, that is simply an inaudible and should be treated as such.
If there is an important, recurring word or name that you can't verify, but which we really shouldn't allow to go through to delivery as-is, submit the edit without certifying (see below for details) and leave a comment explaining the situation.
Part of your job as editor is to verify that the TC grades accurately reflect the quality of work and, if not, to regrade them using the tool in the "Fix Transcript Grades" box on the edit assignment page.
Graders spot-check, and the system screens for certain issues, but as the editor you listen to the entire audio, so you are the only one in a position to truly verify that the TC grades are appropriate.
The hope is that most TCs will reach the edit stage with appropriate grades, but regrading where needed is an important part of the editing process, and is not optional. You may adjust grades either up or down, if the quality of the work does not match the grade.
Splices are the notes that separate each TC in the text you are editing. They usually look something like this:
--------------------SPLICE TCID:XXXXXX min:5.15.00-10.15.00 Grade 7/9 --------------------
In most cases, there is also a 15-second overlap between TCs, so you will usually find a line or two of matching text at the end of one TC and the beginning of the next.
You may want to leave the splices in while you're working, so you can tell where TCs begin and end, and what grades they have, but the splices and overlap must be removed before you submit, so the whole document reads smoothly.
Leaving splices in a submitted edit is cause for rejection, no matter how good the rest of your work is. The only exception is detailed below, in the section on incorrect or missing TC text.
At the bottom of each edit job, there is a space for you to write your comments, and a checkbox labeled, "I certify that this transcript is good enough to deliver to the customer."
As the final editor, your job is to submit customer-ready work, and you should be able to do so in most cases.
Wanting your work to have a "second set of eyes" is not generally reason enough to skip certifying. If someone else needs to edit your work, then you should return it so someone else can do the whole job.
In some rare cases of extremely problematic jobs we may authorize you to submit partially completed work, or a job that needs a whole second edit, but only with explicit permission from Support beforehand.
Submitting incomplete work, even if not certified, is cause for rejection.
The presence of inaudibles does not mean a transcript can't be delivered. If both you and the transcriber have listened carefully to the inaudible spots and truly can't tell what's being said, a timestamped [inaudible] tag is fine.
If there was a bad or missing TC but you corrected or transcribed it, so the transcript is now complete and seamless, go ahead and certify. Be aware that your transcription work will not have anyone else grading or editing it, so take extra care with it, and leave it in "final edit" condition, with timestamped [inaudible] tags rather than [xx].
If the customer orders timestamps, these will be applied in a separate step once the final edit is complete and deliverable. For the most part, you won't know whether a given edit is going to receive timestamps or not. Your role is editing (and timestamping inaudibles), so you should certify as deliverable even if the customer notes mention timestamps and the transcript does not yet contain them.
If there is something wrong with the transcript that is beyond your control, then do not certify as deliverable, and explain in comments why you could not.
Bad or missing TC
If one of the TCs was missing or contained the wrong text and you did not or could not fix it, follow the instructions in the "Special Situations" section for marking the problem spot, complete the rest of the edit, and do not certify.
Something's wrong or weird
If something is wrong with the audio, the customer left instructions that can't be followed or don't make sense, or something else needs attention from CW before the transcript is delivered, leave an explanation in comments and do not certify.
If you don't leave a comment to explain why the job is not deliverable, your work will be rejected no matter how good the edit may be.
See the Main Style Guide for full details on following customer instructions, and what to do if the instructions contradict the Style Guide.
Customer comments may say something like, "Skip the first 6 minutes. Just start when the interview begins."
The transcribers are instructed to include all content in their TCs, because they can't always tell which part of the audio they have. As the editor, you must trim the audio as requested. Simply delete any transcribed material that falls outside the portion specified in the special instructions.
Customer comments may say something like, "No need to transcribe the interviewer. I just need the responses."
If the customer asks that a particular speaker be omitted, then that person's words should not be transcribed. If a transcriber included them, they should be deleted.
However, unless the customer specifically instructs otherwise, the transcript must include a tag to indicate that someone was speaking. Choose a simple tag such as [interviewer speaks], and use it consistently, without a speaker label, in place of the actual speech.
Chris: That's what I thought.
Chris: That's an interesting question...
See the Main Style Guide for information on tagging with the language name or [non-English speech].
As the editor, you need to make sure foreign language is treated consistently within any given transcript, using only [non-English speech] or only a tag with the name of the language, not a mixture of both.
If you identify the language in your tag, use the language name only, without adding words such as "speaks." In other words, your tag should be [French], not [speaks French] or [speaking French].
Occasionally, a transcriber may submit the wrong text for a job, or transcribe only part of the audio in a TC. If the work is auto-graded, this might not be caught until edit.
If you discover the problem before investing too much work in the assignment, you can use the "Report a Problem with this Hit" link to alert Support, then return the job.
If you have already put a lot of work into it, you can go ahead and complete the edit, as long as you do one of the following:
If you want to transcribe the missing content yourself, and you have enough transcription experience to do a good job of it, then by all means go ahead and fill in the missing TC (or portion of a TC).
The difference between being a Expert Editor and earning Master Editor status may come down to whether you are willing and able to transcribe any missing or incorrect text in an edit, especially a short stretch of a minute or less.
Your work will not be graded or edited by anyone else, so make sure to leave it in "final edit" condition, with timestamped [inaudible] tags rather than [xx], etc. Remove all splices and certify the edit as ready to deliver, assuming everything else is OK.
Downgrade the problem TC as appropriate, and email Support@CastingWords.com so we can look into the problem TC, and also pay you for the extra work.
If you are not a skilled transcriber, or simply don't want to do the missing transcription, you may instead:
If you don't transcribe the missing content, and you've put too much work into the edit to return it, you may mark the bad spot as follows:
Leave the splices in and tag the bad spot
Leave the splices before and after the bad, incomplete, or missing TC in your edit. Delete the TC text and replace it with a tag indicating that text is incorrect, incomplete, or missing, and that the TC needs to be re-transcribed.
--------------------SPLICE TCID:XXXXXX min:5.15.00-10.15.00 Grade 7/9 --------------------
[incorrect text -- TC needs to be re-transcribed]
--------------------SPLICE TCID:XXXXXX min:10.00.00-18.00.00 Grade 9/9 --------------------
Downgrade the bad TC(s) and leave a comment
The bad TC should have been rejected in the first place, so you should use the regrade tool to change the grade to 4.
Edit the rest of the document, leave a comment about the incorrect or missing missing TC, and submit your work without certifying.
Even if part of a TC is present, if you are not going to transcribe the missing content yourself, the whole TC should be treated as bad. If part of a TC is unusable and you don't fix it, the whole TC must be redone. Delete and tag it as described above.
It costs us more time and money to re-transcribe a minute or two and insert it into the transcription than to re-transcribe the whole TC. Do not edit a partial TC and just tag or timestamp the missing spot.
Editors are expected to submit customer-ready work in most cases. While we use the same 0-9 grading scale as other types of assignments, we consider edit grades to be more pass/fail than other jobs.
If your work is not a customer-ready 8 or 9, or a solid 7 with little to no intervention needed, it will be rejected.
Any one of the following issues is cause for rejection, regardless of how good the rest of your work is:
Failure to spell check
[xx], [?], or [sp] left in
Splices left in (except in cases of a bad TC)
[inaudible] or [indecipherable] tags not timestamped
Timestamps used anywhere but [inaudible] or [indecipherable] tags
Most speaker label problems
Incomplete work submitted
Edit not certified, and no explanation is given