June 30, 2005

Bringing journalism into the 21st century

In response to my posts on the liberties that reporters take with quotations from interviews, Linda Seebach sent a persuasive email in defense of her profession. I'll quote her in full, and then argue back.

She wrote:

But most reporters don't have transcripts; we have notebooks.

So the person I'm interviewing comes out with a killer quote, and I write it down with great care. But writing is much slower than talking, so by the time I've got that sentence down, he's 150 words further on.

When I get back to my desk, and I'm writing about the interview, I assume as a practical matter that any two successive remarks were separated by a lot of stuff I didn't or couldn't get down, and I don't run them together as if they were a single sentence. As long as I don't misrepresent the sense of what the person said, though, he doesn't have any more exact memory of what was said in what order than I do.

We ran a feature for a while in which we interviewed interesting people, and then edited the transcript of what they said "for length and clarity" as we said on the page. The editing involved very substantial revisions, in fact. Not only for length -- from 6,000 words or so in a half-hour interview down to about 1,800 -- but also in order of topics and putting related material in one spot even if the speaker mentioned it at different times. People ramble. Many of the people so interviewed said they liked how the page came out, but not one ever so much as mentioned how extensively the transcript was edited. I doubt they noticed.

As Lord Peter Wimsey's Harriet observed, after Lord Peter's valet had repacked her luggage, she was surprised to find herself such a neat packer.

Let me note in passing that my own experience of being revised by journalists has often been less positive. I notice, acutely and painfully, when things are taken out of context or rephrased so as to mean something new and unintended. I won't whine about my own experiences, but I've documented several other people's in some detail, for example here and here.

It's a very real problem that typical speech rates are around 150 word per minute, while typical handwriting speeds are more like 20 wpm. That's why British journalists are still required to learn shorthand (though the certification level is only 100 wpm). Rex Rhoades discusses the problem from an American perspective ("Should reporters learn shorthand for notes?"):

Either people need to speak more slowly or I need to write more quickly. I’ve been telling myself that for 25 years and neither one has happened yet.

I mourn the great quotes I’ve lost to indecipherable notes and the quotes abandoned when speakers simply outran my note-taking ability.

And it’s not just me being left in the dust by fast talkers. When I peer over a reporter’s shoulder at some freshly recorded notes, the mess I see never inspires confidence.

Yes, a tape recorder can help. But try juggling tapes for three multi-source stories, all of which are due in two hours. The benefit of the recorder is usually sacrificed to the reality of the daily writing load.

Thus the question I recently posed to a variety of journalists and journalism educators here and abroad: If taking notes is so important, shouldn’t journalists know shorthand?

There’s a world of opinion out there — literally. While U.S. educators are generally cool to the idea, the skill is considered indispensable to print reporters on the other side of the Atlantic.

Bernie Corbett, national organizer for Britain’s National Union of Journalists, explained that most journalism jobs there require an academic “qualification.” And, he said, most print journalism degree programs require shorthand skills of 100 words per minute.

But shorthand is the 19th-century solution. This is one of those cases where 21st-century technology really can help. The existing technology -- with a bit of training and effort -- can solve most instances of this problem fairly easily. The technology of the (near) future should make it nearly trivial.

With existing technology, you can easily upload recordings -- from an iPod, or a minidisc recorder, or an old-fashioned analog recorder, or whatever -- to your laptop. You can let the uploading run while you're doing other things. If the interview circumstances permit, you can record from a microphone (or a telephone interface) directly onto your computer. With a bit of cleverness, you could do direct recording via wireless remote, using this (cellphone to cellphone to PC) trick, or some other way of controlling digital audio recording on your computer via a wireless link. Perhaps there's already a product out there that will do this out of the box, with no cleverness required at all. If it's not true now, then it won't be long before wireless handheld devices combine intercom-like ambient audio I/O with remote control of transfer to a computer suitable for transcription and editing. Or maybe you'll be able to access your handheld's mass storage without any transfer.

Once you've got access to the audio on a real computer, you can use convenient free software (I use Transcriber) to create (typically partial) transcriptions. By moving through the file in a nonlinear way, you can find and mark the crucial bits, and transcribe them or make a note for more detailed examination later on. No "juggling" of multiple cassettes is required -- you just open as many windows as you need. You can cut and paste between the transcript and your notes or your story.

If I were a reporter, that's how I'd do it. In fact, even though I'm not a reporter, that's how I do it anyhow, for those Language Log posts where there's a question about who said what when. The biggest pain is that I usually need to get the audio from some streaming source like CSPAN or CNN or NPR. And don't tell me about deadlines, please -- my general policy is not to spend more than half an hour on a post, and I usually manage to stick to that. As I mentioned in my post on Ipsissima Vox Rasheedi, getting an accurate transcript in that case took me about 12 minutes, of which about 10 minutes was spend finding the right part of the (streaming) interview.

In the near future, you'll be able to use automated media indexing on your audio archives to make it easier to find the bits you want. In this Language Log post, I discuss my experiences with an existing HP system called SpeechBot. In my opinion, these applications are now probably not worth the trouble. (At least this is true for the applications we're talking about, where you're dealing with a few hours of audio that you've heard at least once -- the situation is different if you're looking for something in hundreds of thousands of hours of otherwise unindexed material). However, the basic enabling technologies (figuring out who spoke when, and turning speech into text) are improving steadily, and within a few years they should be good enough to be useful for interactive audio search in many applications, including this one.

I should also note here that more and more public discourse -- certainly press conferences and speeches and hearing and so on -- is recorded and archived anyhow, so that reporters don't in principle need to make their own recordings at all. In many cases, a real-time subtitle-quality (or court-reporter-like) transcript is also created. Right now, there is probably an issue about delay in access to such material, which makes its use impractical for reporters on deadline. But there's no reason in principle that such recordings could not be made available as a real-time feed, spooled onto reporters' laptops while they make notes to themselves about less objective things than who said exactly what exactly when.

It's also easy to imagine software for real-time annotation of a recording -- in the simplest case, a reporter could just hit a button that would mean "what (s)he just said was important, flag it." Adding a few words of comment via stylus or keypad (or keyboard) would be just as easy.

Reporters ought to be leading this transformation. I mean, even when quotations from sources are largely ceremonial in character, why not make them accurate? What do you want to bet, though, that bloggers will be using such techniques routinely for years before journalists take them up?

Posted by Mark Liberman at June 30, 2005 10:12 AM