Managing virtual assistants: the surprising benefits of transcription
| speakingI frequently give speeches. During some months, practically every week involves a presentation or two. I usually post presentation, recording, and notes for these presentations, but it would be handy to have a transcript. Timestamped transcripts also make it easy to search within presentations, synchronize audio with slides, and even remove ums and ahs.
I’m not an auditory learner. I find it difficult to sit still for an audio-only session, even if it’s my own. ;) I’ve transcribed some things–my research interviews, a few of my talks–using the handy Transcriber program, which made it easy for me to associate text with specific audio segments.
And maybe transcripts can help me learn how to be a better speaker, too! I speak at about 200 words per minute when I’m excited. While that’s below the 300 words per minute I often joke about, it’s still well above the recommended 140-160 words per minute for persuasive speeches. Transcripts make my rambling sentence structure and my verbal crutches painfully obvious, too. ;)
If I can get word counts and review what I’m saying without the large initial effort of transcribing things myself, I think it’ll be well worth it. It gives me metrics, and metrics are useful. Like the way that people work on getting into a target heart beat zone when exercising, these numbers can help me get into a target speaking rate zone, providing me feedback about going too quickly or too slowly. And like the way that listening to music and practicing on the piano will eventually give me a feel for how long a quarter note is in different tempos, listening to good speeches and practicing myself (either through actual presentations or through podcasts I make on my own) can help me adjust my speaking rate.
On to the actual process of transcription:
I posted a job notice on oDesk looking for people who can edit and transcribe audio files. While waiting for candidates to respond, I asked one of my virtual assistants to download Express Scribe and give it at try – it might help her develop new skills. I like having plenty of timestamps in the transcribed text because it makes it really easy for me to recheck the transcript, so I also sent her a link to this shortcut for timestamping files.
A few good candidates responded to my oDesk ad. One of them had an excellent sample transcript, so I’ve also added her to my growing team. I sent her the audio recording for the talk I did yesterday, and I’m looking forward to getting it back.
Here are some notes on my preliminary experiences with transcription, and I’ll add more as I explore this:
- More effort is required to transcribe ums, ahs, repeated words, sounds, and other things accurately. If you don’t need them in the final transcript, tell your transcriptionist to skip them. If you want to make it easier to edit the file, you can ask people to add timestamps and a marker like “!!!” during the ums and ahs. Work backwards from the end of the file in order to remove the ums and ahs, so that you can keep the timestamps useful for as long as possible.
There’s probably a better way to handle this audio editing. Maybe a transcriptionist could remove ums and ahs along the way? Maybe I can ask an audio person to clean up the audio before handing it over?
- I need to pause more when giving presentations. ;) Pausing more helps transcriptionists figure out sentence punctuation and paragraph separation.
- If there are unclear words, ask the transcriptionist to indicate that with a timestamp and a marker like ???. That way, you can easily review and fix it.
I wonder how I can take advantage of Dragon NaturallySpeaking here, as I already have it. Even better if I could get someone else to train and correct my user model, but I think Dragon NaturallySpeaking wants me to upgrade to the super-expensive version in order to do that. =|
Hmm… How can I tweak this process…
Do you outsource transcription, or do any of your friends outsource transcription? I’d love to hear about experiences and tips!
1 comment
David Ing
2009-04-05T20:53:44ZThanks for the pointer to Transcriber. I see that it's easily installable on Ubuntu, so I may have to try it some time.
I personally don't find that full transcripts are worth the effort. There's a saying in consulting that if you don't do it during the day, you'll be doing it at night. This means that post-production activities (i.e. editing) should be minimized in favour of real-time activities when possible.
When I attend talks, I create text digests, even when I have an audio recorder running. These are valuable (a) for someone who hasn't attended the talk, to judge whether he or she really wants to spend the real time listening to content, and (b) as lightweight indexes in the content.
When I was visiting with colleagues at Research about the merits of text over audio, they pointed out that text has a feature that it's really easy to index. Streaming media aren't so straightforward.
You're in a slightly different situation that me, in that I do make recordings of my own talks, but I don't do transcriptions. (Most of the time when I give talks, I have slides, which serve as an alternative form of index). So the question becomes: are the transcripts for your own personal benefit, or for the benefit of the audience?
You can take some pressure off transcribers -- if you choose to continue to use non-professionals, down from the level of courtroom stenographers mentioned by Paul, above -- if you'll reduce the standard of notation required down to a "text digest", or indexed to the slide numbers. If you do have a digital audio recording, it's easy to use a slider to find a specific point in the talk -- even easier, if the time marks are explicit -- and the effort is considerably reduced.