Amy’s Mailbag: SpeechInk’s nonverbatim style guide, plus a question on accuracy

| June 12, 2013
  • Add to favorites
  • Reddit
  • Facebook
  • Twitter
  • Google Plus
  • Email
  • RSS

In this edition of Mailbag, I’ll be telling you where to find that elusive SpeechInk nonverbatim style guide, plus I’ll answer a question on grading.

The SpeechInk nonverbatim style guide

Says a reader,

“SpeechInk has nonverbatim HITs, but there’s no style guide for them. I looked on their HIT and on their site. What style do you use for those transcripts?”

Glad you asked. SpeechInk does have a nonverbatim style guide. Two, in fact. They’re just absolutely terrible at telling transcribing Turkers where those guides can be found!

What makes their documents even harder to find is that they have two sites, one under the name SpeechInk (which is how they do business on Turk) and one under the name SpeechPad. The two companies are one and the same. Their CEO has explained there are the two names for legal reasons only.

All the company’s style guides are on that Speechpad site.

Need a link? Right this way. I made this post last fall, when I first came across SpeechInk’s nonverbatim style guides and published the link. Just scroll down a little bit on that page to the green text; the link is in the middle there.

The actual link’s already changed once, which is why I’m going to keep providing it through that post rather than here or in a continual series of new posts about it. That way, everyone can always find it in that one place, and I’ll keep it current there.

Clean up or get the actual words?

My next Mailbag question is about a transcript-grading example. I love those, because they can give valuable information about a company’s standards. They also give it in a more real-world way than a style guide; sometimes that’s easier to understand.

This reader writes (I added the bullet points for clarity here)…

“Something that left me confused is on the CastingWords Grade Examples page, example of a 6.

  • The original sentence: ‘Some machines might not complete their job – it may have been switched manually to do something else – so you have to a build-in the [xx] to restart that test elsewhere.’
  • The 9 version is: ‘Some machines might not complete their job. They may have been switched over manually to something else. So you have to have a build-in resilience to restart that part of a task elsewhere.’
  • My version would have been ‘So you have to build in resilience…’ or ‘So you have to have a built-in resilience…’ Would either of those choices be marked as an error?”

Great question! The first thing to know about the CastingWords Grade Examples is that they have not been redone in a while, including not since their current style guide was released. So they shouldn’t be looked to as an example of exact style, but they can be useful for understanding the types of things that go into grading.

The second thing to know is that without any accompanying audio, it’s hard to say whether this reader could have done what they suggest and gotten a great grade. Yes, their changes make more sense of the text. But because this is transcription work, rather than editing work, the real key to whether those new changes would be better or worse is what actually was said in the audio.

On the one hand, if the speaker actually said, “So you have to have a build in…,” then this reader’s first version would be wrong, because it missed words (the second “have” in that last sentence) that were spoken. That isn’t OK, even in nonverbatim transcription. And actually, the second version would be wrong too! Not because it missed words, but because it substituted a wrong word (“built” instead of “build”).

Substituting inaccurate words is just as bad as missing words. For example, it’s possible that “build-in” is a bit of lingo specific to the field being discussed. Also, the client may be wanting to use the transcript to quote the speaker. Finally, even if neither of those is true, if that’s how the person spoke throughout the interview, the client may have wanted some sense of the speaking style of the person in the transcript.

For all those reasons, if “build-in” is what was spoken, it’s safest not to change it. Most any transcription company will not penalize you for using the speaker’s words as they were said.

On the other hand, let’s say that the speaker said, “So you have to have a built in….” If that’s what was said, then my reader’s second version would be fine. They added a hyphen, which is just a grammar judgment call and just fine here. But that first version would still be wrong, because the speaker’s second “have” in that last sentence still would be missing.

I’ve said before that in transcription, style trumps grammar. Thanks to this reader’s question, you can see that accuracy trumps grammar, too!

Want to know more about why each of those snippets was graded the way it was? There’s a brief explanation on CastingWords’ webpage, where you’ll find the original example my reader referred to. I’ve also just finished and posted a members-only piece on transcript grading; it uses those same two snippets as an example, going through each one and pointing out what differences between them justified a lower grade.


Useful? Help others find this, too!

Tags: , , , ,

Category: Amy's Mailbag

Comments (2)

Trackback URL | Comments RSS Feed

  1. Rainbowscape says:

    I haven’t done much for non-verbatim for SpeechInk lately, because I felt like their instructions needed more clarification.

    SpeechInk does a great job offering feedback for their verbatim HITs. However, I would really like to get more feedback on the non-verbatim HITs. I wish the SpeechInk Style Guide would have a sample transcript attached to it.

    For instance, I’ve always put colons behind my speaker labels in HITS with more than one speaker. Examining their Style Guide, is that wrong to do with their non-verbatim transcription HITs?

    • Amy Marre says:

      Agreed, Rainbowscape—they need to do a way better job with feedback on their nonverbatim work. It would be better for their clients and a better experience for those transcribing!

      Based on their style documents, it’s true—there’s no support there for putting colons in with speaker labels in nonverbatim transcripts. I previously did some of their earnings calls work, and there (as indicated in these nonverbatim guides I linked to), you’re just supposed to name the speaker on one line and then start what they said below their name.

      You could try emailing them about this and seeing if they respond. Or, if you’re feeling brave, you could also try not putting any colons in with speaker tags in one of your next nonverbatim transcripts for them and then seeing if your score improves. (I’m pretty sure they won’t reject work from you for changing that factor alone.)

      I have a feeling it either will not affect or will improve your score slightly. If it helps, the one saving grace about SI’s lack of investment in their nonverbatim style is that they don’t grade style issues as strictly for those transcripts. I always hope that’s because they know they have no real right to, considering they don’t give Turkers much help with finding the corresponding style guides!