Intent

This page describes the overall quality expectations of produced transcripts. This page is not intended to go over every detail of formatting which can be found elsewhere (like Processing Standards).

Summary

Our customers require impeccable attention to detail on their transcripts, particularly on information that is critical to the content of the transcript. Examples of critical content include but are not limited to:

• Provided spellings and consistent spellings (whether spelled or unspelled)
• Numerical information, including correctly scribing ambiguous or vague numerical amounts
• Reasonably following customer instructions (for example missing the only directive in a job is probably a big deal; missing one out of multiple instructions in a job is comparatively less serious.)
• Obvious errors in readability or accuracy
• Other information that if incorrect would vastly change the meaning of the overall job
• Anything that could make us look bad

If you notice that any of these listed errors are present in your review jobs, it's important that you take that feedback into account to modify your habits while working such that these errors do not happen.

Metrics We Use

For editing work (that is, when editing work from VR or from human scribes), the primary metric we use is whether there were any instances of errors with respect to critical information. If there are any of these kinds of errors to be found in recent transcripts, employee work is considered to be unacceptable, the job in question would be marked as FDS/FQA. A single critical error and/or critical errors in one job will not automatically disqualify an employee from incentives. The goal here is NOT to penalize people for errors but instead to establish the necessary expectation so that customers receive consistently acceptable quality work.

These five categories of information must always be correct:

QAs - these five categories of information also count as reasons for FDS/FQA (see the bottom of this document for more info)

1. Provided spellings and consistent spellings (whether spelled or unspelled)

Any time a provided spelling is given we need to capture this accurately.

• A name spelled at any point in the transcript needs to be accurate and consistently spelled throughout (e.g. "John that's J O N" might prompt ctrl+F and search and replace for John --> Jon)
• The same name "Cate" should not be later transcribed as "Kate", "Cait", or others.
  • VR cannot be counted on to be consistent with these spellings. If you change one spelling, consider using ctrl+F to change en masse.

You can view the Critical Errors page for additional information.

2. Numerical information

Any time a numerical amount is dictated we need to capture this accurately. Note that in some instances the correct way to accurately capture the dictation is to scribe verbatim, e.g.:

• Spoken: "They have a couple thousand dollars" should be transcribed as "They have a couple thousand dollars" and not with numbers.
• Spoken: "They have a hundred and some thousand dollars" should be transcribed so that when read aloud, the text matches the audio, e.g.:
  • OK: "They have 100 and some thousand dollars"
  • OK: "They have a hundred and some thousand dollars"

You can view the Critical Errors page for additional information.

3. Reasonably following customer instructions

We understand that some instructions require discretion and can be interpreted multiple different ways. Still, these should be followed to the best of our ability.

• Directives for subject and text should generally be followed instead of transcribed in the text
• Similarly, other directives should be followed as well (period, comma, new line, bullet points, etc.)
• Spoken: "Had my meeting with Benny B E N N I E" should be transcribed "Had my meeting with Bennie"

You can view the Critical Errors page for additional information.

4. Readability or Accuracy errors, particularly errors which are obviously discernable simply by reading the text

We receive conversational English which often is spoken without strict adherence to grammatical standards or conventions. We wish to grant staff sufficient leeway to use discretion on how to transcribe appropriately. Here are examples of objectively poor transcriptions that were not transcribed appropriately:

• "Met with them for an annual review this is the first Monday in January at 2:00pm we reviewed asset allocation. Performance. Risk tolerance. Time horizon."
• Spoken "She sold her stocks" transcribed as "She sell her stocks."
  • Although the meaning is potentially preserved, anyone reading the transcript would immediately know this was wrong.

You can view the Critical Errors page for additional information.

5. Other critical information and anything that makes us look bad

• This transcript is egregiously wrong. Assume the underlined text is omitted: "They have three kids. We need to do a $30,000 policy for each of them.
• Copytalk should be scribed as such, not "CopyTalk" or "Copy Talk" or "copy talk"
• VR hallucinations or omissions causing the text to be inaccurate must be fixed before sending the job out.
• Although every error could be construed as "making us look bad", we are most concerned with egregious errors that are unprofessional or grossly inaccurate. (for example TDAmeritrade, though formatted incorrectly, is not an egregious error as it is at least formatted as a proper noun). Also, we outline "Less Important" errors below which, although potentially making us look bad, are also not generally considered egregious errors.

You can view the Critical Errors page for additional information.

To Keep In Mind When Editing VR Work

A portion of VR errors are obvious and require little extra attention to fix. But many errors are more subtle. While we expect that someone new to editing work will miss errors (we use the term "FTC" or "failure to correct"), by the time someone has about a dozen shifts spent editing work, there need to be no FTC errors on important information.

Here are some examples of subtle-yet-important-to-correct errors the VR may commit:

• Spoken "We touched on personal life changes" transcribed as "We touched on personnel life changes"
• Spoken "forty one thousand" transcribed as "41 1000" instead of "41,000"
• Spoken "two hundred and thirty thousand dollars" transcribed as "230 or so dollars" instead of "$230,000"
• Spoken "she retired from Food Lion" transcribed as "she retired from food lying"
• Spoken "Benny B E N N I E" transcribed as "Benny"
• Spoken "asset allocation" transcribed as "S allocation"
• Spoken "required minimum distribution" transcribed as "requirement for distribution"

All of the above represent critical errors which should never happen in a transcript. We encourage the use of search and replace (Ctrl+F) when making changes. If the VR put "Benny" instead of "Bennie" once in a job, don't rely on your focus to catch it! Institute a habit (search and replace) to fix them all at once when the issue is first found, and have an easier time in the rest of the job

Examples of Less Important Errors

While we do expect perfection on some categories of information, we do not expect absolute perfection or adherence to Common Terms/Standards in every situation. It's important that as a transcriptionist you are able to exercise discretion to make the correct choice to deliver the best transcript possible in a reasonable amount of time.

Formatting of jargon

Often, financial jargon has multiple acceptable spellings; for example both 401k and 401(k) are acceptable. In this and any other situation, pick the format to use within the transcript and stay consistent. Failure to consistently transcribe the identical term detracts from the quality of the job.

Minor punctuation issues

Transcripts should follow basic English composition rules. That said, when we are delivering work to a narrow audience (one client and/or one client and their team), we are less concerned with adherence to formal grammar standards as we are with overall accuracy and efficiency.

For example, while commas should be used when two independent clauses are separated by a conjunction, and while the ideal transcript uses commas appropriately, failing to use a comma in this situation is not considered itself an egregious error:

• Spoken: "We went over the plan from last year and we discussed our thought process for this year's plan."
• Transcribed with minor error: "We went over the plan from last year and we discussed our thought process for this year's plan."
• Ideal transcript: "We went over the plan from last year, and we discussed our thought process for this year's plan."

Similarly, while commas should not be used when two independent clauses are spoken without a conjunction, and while the ideal transcript does not use what's known as a "comma splice", using a comma in this situation is not considered itself an egregious error. It is considered somewhat more egregious than failing to use a comma, however.

• Spoken: "We went over the plan from last year. We discussed our thought process for this year's plan."
• Transcribed with minor error: "We went over the plan from last year, we discussed our thought process for this year's plan."
• Better transcript: "We went over the plan from last year. We discussed our thought process for this year's plan."

Transcribing incorrect grammar

Often customers dictate incomplete sentences or dependent clauses as standalone sentences. Transcribing text that way is fine if that's how the customer said it. The transcripts we produce do not need to overly focus on issues like this. Customers speak in conversational English and even when not grammatically correct, our transcripts can often reflect conversational English.

• Spoken: "They said that everything is going well. (clear pause). Exception being, Sandy had broken her arm."
• Transcribed (acceptable): "They said that everything is going well. Exception being, Sandy had broken her arm."
• Transcribed (acceptable): "They said that everything is going well, exception being Sandy had broken her arm."

• Spoken: "They have several friends who have gotten into this age-restricted community. Let's see. (clear pause). One in their 70s, one in their 80s, and one about to turn 65."
• Transcribed (acceptable): "They have several friends who have gotten into this age-restricted community. Let's see: one in their 70s, one in their 80s, and one about to turn 65."
• Transcribed (acceptable): "They have several friends who have gotten into this age-restricted community. Let's see. One in their 70s, one in their 80s, and one about to turn 65."

Inferred units of measurement

We want to be cognizant of the patterns and quirks of Copytalk Voice Recognition (VR) and ensure that the use of Copytalk VR is as a whole actually saving time and effort from staff compared to transcribing from scratch. VR often infers a unit of measurement. For example:

• Spoken: "They have one hundred thousand dollars in their Raymond James account and fifty in their account here."
• Transcribed by VR: "They have $100,000 in their Raymond James account and $50,000 in their account here."

In this example, although the spoken audio does not literally say "Fifty thousand dollars", using context we can confidently validate that the VR transcript is correct. While keeping the transcript accurate as to the spoken audio is nearly always acceptable, we do not wish to mandate that staff change correct inferences by the VR just for the sake of having a close-to-verbatim style of transcription. We transcribe near verbatim, not verbatim, and we want folks to be able to process this work efficiently and without an undue burden.

That said, the VR commits errors and it's critical that the units of measurement are correct!

Inferred or Guessed Spelling

Names are most often dictated without being spelled. Names often have multiple legitimate spellings. While consistency with a spelled name is important, if a spelling is not provided we recognize the need to use discretion when choosing a spelling.

In general, the prevailing or most common spelling of a name should be used. For example, if a customer says "Carol", scribing this as "Karyle" would be unacceptable. Yes, there is at least one "Carol" who spells their name "Karyle" but it is not the prevailing spelling. In this instance, "Carol" is probably the best choice.

We want to recognize there are multiple appropriate ways of spelling names. For example: Tommy, Tommie, Tommi are probably all reasonable choices (possibly using "Tommi" for a female-identifying individual), but "Tommey", "Tawmy" etc., should not be used unless that particular spelling is given.

We do not wish to specify an exhaustive list of mandatory spellings and instead wish to give staff discretion on how unspelled names should be scribed.

Other Minor Errors

While we wish to restate that, after editing a transcript, the transcript should be basically perfect, we do recognize that some people will be able to hear sections better than others. Some folks may leave a section with a QM or Garbled bookmark and others will be able to put the correct text. We expect some slight variation between people and as long as the variation from the dictation is minimal, and all important information is correct, that is likely good enough for our purposes.

Changes That Do Not Require Your Time

There are often multiple correct ways to transcribe something. Examples include but are not limited to:

• 401k and 401(k)
• e-mail and email
  • and other words which are sometimes hyphenated and sometimes not (e.g. co-operate and cooperate)
• et cetera and etc.

Although the transcript which you produce is what you are responsible for, note that if the VR chooses formatting different than your preference, and it is not something that needs to be specifically formatted per Common Terms or Standards, that you do not have to edit these.

What About CPH and FDS%?

The CPH and FDS% metrics listed on the compensation policy were primarily designed with scribe work in mind. When editing work (that is, not having to originate and create the entire transcript yourself), it's expected that there will be drastically fewer errors.

We do not recommend using CPH and FDS% metrics for gauging your quality in editing work. Your primary indicator will be the jobs you receive back for review. If you notice that there are usually changes from QAQA, this likely represents a deficiency in the quality of your editing work. Take some time to develop new habits! Even something simple like pausing at every provided spelling or number to do a "focus check" can result in improvements.

Reasons for an FDS (False Direct Send) and FQA

When QAs determine that a job does not meet the minimum quality standards, the job will be marked as FDS (if the prior employee sent the job direct to the customer) or FQA (if the employee requested QA review).

A note on what constitutes FQA: One of the QA's responsibilities is to provide quality and training by assisting with difficult dictations. They ARE NOT there to fix what should have been done by the preceding employee before sending the job out. Jobs sent to QA are expected to have sections of text that may not match up with the audio. However, jobs sent to QA must still be free from mistakes independent of the job's audio and dictation quality (Standards, Common Terms, reasonably clear spellings, reasonably clear instructions, etc.) Even if sections of text are bookmarked and, in the discretion of the QA, determined to be reasonably clear such that the text should have been correct before sending it out, the job can be marked as an FQA depending on the severity of the error.

Errors Which Constitute "Automatic" FDS or FQA

The presence of a single "critical error", defined in the beginning of this page, is reason for FDS/FQA. The minimum acceptable level of quality is for transcripts to be free from critical errors. Of course we prefer no errors at all but we recognize that minor errors are often just that: minor.

QA FDS

For QAQAs: Errors of any of the categories above should be marked as Problem Fields, and the job should be marked as an FDS if the QA failed to correct a critical error. It's extremely crucial not to take the QA's iteration for granted as it can still contain serious errors that the QA missed.

Further Examples For QA Use

You can reference the Critical Errors page for additional examples.