My G+ Profile
« $100 Laptop Design is Firming Up | Main | Forget the Frontal Lobe »
Podcast Transcripts Via CastingWords
Jon Udell is transcribing his podcasts using CastingWords. (Also see Jon’s earlier review of the service.) His bill? $260 for 620 minutes of audio. That’s darn cheap.
Jon reports that the results are surprisingly accurate, but he edits them to make people read better than they sound and to make them more readable. This process takes an hour or so per episode—almost one-to-one on a time scale. Of course Jon’s also spent time editing the audio for the podcast as well. That’s a considerable time investment, but the results show quality.
Why do this? One answer is that Google indexes written text, not audio, so getting search placement requires text. Another reason is that some people simply prefer to read instead of listen.
Doug Kaye used to create transcripts of IT Conversations shows back when it was mostly interviews he was doing. He stopped, mostly because of the cost, I think. We create fairly extensive landing pages for shows that should contain significant textual content to help Google index the shows correctly. Still, there’s no way you can completely capture the multiple twists and turns of a standard show in a page of text.
Posted by windley on August 15, 2006 7:11 PM
Comment from Larry Press at August 17, 2006 11:50 AM
Casting Words says they use human transcribers. Are there services or programs that do speech recognition on mp3 or other audio files? If so, how is their accuracy?
Comment from wade at August 17, 2006 2:00 PM
Take a look at http://www.podzinger.com/ they are building the Google for video and audio just for the reason search is largely based on text.
Comment from Choong Yong, KOH at August 17, 2006 8:49 PM
I personally think that transcripts are still better than the descriptions on the landing pages. Podcasts are good for first-time consumption of the information, listening insteading of reading. However, when one wants to refer back to certain portions of the podcast where some interesting point was made, it is usually not easy to look for the correct point of time.
I believer human transcribers still produce the best quality. Maybe some thoughts can be channelled to trying have podcasts divided into digestible chunks (maybe 5 minutes) and volunteers can pick up the chunks to do human transcripting. An editor can do the final editing before the transcript is published.
Comment from Parag Shah at August 21, 2006 4:15 AM
People usually listen to podcasts when they are driving, working out, walking, etc, when the transcript would not be very useful. Sometimes I do want to just give the material a quick read if I do not have time to listen to the entire audio, but that's very rare.
Transcripts as you rightly mention is very useful, if we want to search for a relevant podcast. Also a feature that I think may be useful is a transcript that has pointers to the relevant section in the audio. So if a podcast has three sections, then an entire transcript with links to the sections (such that we can directly hear that section when we click on the link) will be very usefull. That way people can find out what sections they would like to hear in a long podcast.
Comment from Russ Cochran at August 24, 2006 2:49 PM
I am looking for a digital recorder which can be used to record an interview, then take the disc and input into a computer to make a written word document of the interview. Can you help?
Comment from Russ Cochran at August 24, 2006 2:53 PM
I am looking for a digital recorder which can be used to record an interview, then take the disc and input into a computer to make a written word document of the interview. Can you help?
Leave a comment
I encourage you to leave a comment below. Your email address will not be displayed on Technometria, but allows me to communicate with you directly. Your email address won't be displayed, but will be used to compute a MicroID for your comment.