A friend recommended a technical talk today: How to Design an a Good API and Why it Matters by Joshua Bloch. Looks good! It's also an hour long...

For a variety of reasons, it's hard to watch an hour-long video. I'd prefer to read the same content. But it isn't available textually. For my own talks, I produce full text as part of the preparation (for example, the Unicode sandwich talk).

I've even transcribed other peoples' PyCon talks: Stop Mocking, Start Testing, and Speedily Practical Large-Scale Tests. It was a good way to ensure I actually watched them!

People put slide decks up on SlideShare, but decks vary wildly in how well they contain the content. Some simply provide a backdrop, which is entertaining during a talk, but useless afterward.

Is there some way we can pool efforts to get more talks transcribed or summarized? Surely others would like to see it done? And there must be people eager to contribute in some way who could spend the time? Does something like this already exist?

I know the full talk, with the real speaker really speaking to me, is the best way to get their message. For example, Richard Feynman's series The Character of Physical Law just wouldn't be the same without his accent and delivery. But if the choice is reading a lengthy summary or not getting the message at all, I'll definitely take the summary.

Or maybe I'm an old codger stuck in text-world while all the younguns just want video?

» 7 reactions

Comments

[gravatar]
Steve 7:28 PM on 2 Sep 2014

I think you and me both are old codgers. I find the same when people tell me that so-and-so has a review of some item of media, and it's this guy droning at the camera for half an hour when I could read the same as text (with embedded pictures as needed) in a few minutes.

BTW, your www field parser doesn't like g+ URLs with a /+UserName in them.

[gravatar]
Michael Kohne 7:56 PM on 2 Sep 2014

I'm with you all the way. Lots of folks seem very enamored of video these days, but most of it could be done better and quicker via text, in my mind.

One way to approach this (if you had a little money and a little time) would be to try chunking up the video and throwing it at amazon's mechanical turk. Someone would still need to review it, but if you had more than one worker do each segment, you could start by diffing the results of the different workers.

You still need someone to do a last pass on it, but it's a way to start.

[gravatar]
Dirkjan Ochtman 8:35 PM on 2 Sep 2014

Completely agree, and I don't think I can be counted as an old codger just yet. Reading through text allows me to (a) take in the content faster, (b) scan through the content to quickly find the bits I find most interesting, (c) take in the content in environments where audio is impractical (when I don't have headphones) or (d) when bandwidth is scarce.

[gravatar]
Matt Doar 8:50 PM on 2 Sep 2014

It's all about information density and transfer rate for me. Videos captivate well, but I'm unlikely to spend an hour just watching anything at work. Slide decks are pretty low density too. I usually read a paper related to a talk if there is one, or a summary and the readers' comments.

[gravatar]
Jonathan Hartley 12:39 PM on 3 Sep 2014

...agreed all over, plus text is searchable (in-page) & indexable (by Google.)

[gravatar]
Kevin Edwards 5:56 AM on 5 Sep 2014

Yay for text! Moreover, I'd prefer an outline over a linear stream of text.

MTurk is one way to go, but in this case your YouTube video already has subtitles/closed captioning. YouTube can do Automatic Speech Recognition (ASR) which isn't always good, but in this case the subtitles are so good (with punctuation and attribution) that a human must have created them.

You can use Google2SRT to download the subtitles and then hack on them with pysrt. You could even output html that links to the exact location in the video where the line is said.

[gravatar]
Ben Poole 4:06 PM on 6 Sep 2014

Well if you’re an old codger, so am I—transcripts would be excellent for stuff like this, I agree.

Add a comment:

name
email
Ignore this:
not displayed and no spam.
Leave this empty:
www
not searched.
 
Name and either email or www are required.
Don't put anything here:
Leave this empty:
URLs auto-link and some tags are allowed: <a><b><i><p><br><pre>.