Having written up some minutes myself on occasion (in Ypsi and elsewhere), I'm pessimistic about the value of speech recognition software for this purpose. Meetings tend to be noisy enough, between audience noise, the poor acoustics of public buildings, and side conversations amongst the people being recorded, that turning audio into text often requires a human familiar with the subject material and voices, if not somebody who was actually at the meeting.

I don't know any of the details of Ypsi's Comcast franchise, so I have no idea whether Comcast pays the City enough to make tv production feasible. I do lament the fact that I can watch Ann Arbor's Parks Commission and Greenway Task Force and Historic District Commission on tv, but can't get my own City Council; I just don't have any ability to say whether "let's get the meetings on tv" is a reasonable goal.

well, at least more people could/would watch it, were it on cable, so I think it's more reasonable a goal than streaming multiple gigabytes worth of video to the few computers that have the ram/memory/connection to handle it.

Speech recognition will undoubtedly get better with time, I just don't think it's at that point just yet, especially as applied to a noisy little room like the city council chambers. What does that $25,000 we'll be saving pay for, precisely? Is this the salary of someone who only does transcriptions, or is it the time a person spends on that task multiplied by some hourly rate?

Sure, if the goal is to achieve maximally broad access, then cable tv is the way to go. (Or radio...) The advantages I see of an online version are, first, that it's more "on demand", such that those inclined/technically capable can make better use, and, second, that it's free, at least in the model that Steve is using. (Which, I can't resist pointing out, falls into the old saying about open source software - it's only free if your time has no value.)

I assume the $25k is a calculation of hourly rate - there's nobody at the City (to my knowledge) who does anything close to just transcriptions, or even transcriptions half-time (about what $25k would be, including some portion of benefits). That may be an estimate spread across all the entities for which minutes are kept?

I just think that "those inclined/technically capable" are a much, much smaller group than the developers are considering.

As for it being 'free', I'm very hesitant to link any of my own projects directly into a google database, and avoid even using 'google maps' when posting location information. Nobody can honestly say, for certain, what google might do in the next few years, and I think it's myopic for people to sink all their eggs into such a basket.

Murph wrote: "Having written up some minutes myself on occasion (in Ypsi and elsewhere), I'm pessimistic about the value of speech recognition software for this purpose."

If you run speech recognition software over the raw audio of a multi-person conversation, I think your pessimism is warranted. To achieve high accuracy for the basic sound-to-word conversion, speech recognition software requires 'training' for the particular voice it's analyzing. Plus, as already noted, human intervention is probably required at some level anyway, just to tag the transcript with speaker turns and whatnot.

But after training, state-of-the-art speech recognition on its trained voice already delivered mind-blowing results six years ago (natural speed talk even by fast talkers got puked onto the screen almost instantaneously). The limiting reagent for accuracy was processor speed and RAM. 200Mghz and 64M of RAM was sufficient to get very impressive results.

At that time, I used a speech recognition package called DragonSoft to process conversations with two speakers (I was one of them) using the following technique: pipe the digital recording of the conversation into an ear bud; re-speak the words into the microphone that's getting fed to the speech recognition engine, tagging the transcript as you go, pausing the feed to the ear bud as necessary. Otherwise put, use the speech recognition software as a typist.

The cost in time to get a verbatim transcript using this technique, based on my experience, is still going to be somewhere around twice real time on the quick side of things. But if we're going to throw technology at this challenge, I'd rather see technology applied to getting TEXT as the end product.

I think commenters who've mentioned the idea of timestamping videos or loading time-coded partial transcripts to include with audio/video have exposed the fundamental weakness (w.r.t. analysis and research) of all media that depends on real-time consumption: real time. Text does not suffer from this weakness. You can consume text, through reading, at several times the rate that text was spoken in real life. Using very minimal computing power, text can also be analyzed in all kinds of interesting ways (e.g., how many times has John Hieftje used the word 'transparency' at the council table? in what contexts? how does CM x begin a typical comment? how has the discourse on Council changed since Ron Suarez join Council?)

I see the efforts devoted to podcasing, or video webcasting public meetings as interesting exercises in proof of concept, but frankly I see only a small benefit to the community in terms of making the workings of government more accessible or more transparent or in encouraging people who aren't currently participating to do so. Yes, some small number of people might watch/listen-to a council meeting via iPod on on a webcast who otherwise would not. I suspect, however, that those who take advantage of a podcast or a webcast are people who are simply changing from live attendance or CTN viewing to this new alternative medium. I just don't see the potential of looping in new bodies.

There's folks who talk about the impossibility of replacing energy with technology. Democracy is another thing you can't replace with technology. If it's easy, if it's automated, if it doesn't cost something, then I'm skeptical it's offering anything of value. Democracy is a pain in the ass. If it feels effortless, then it's probably not democracy.

Ypsi's meetings are not presently broadcast on CTN, or anywhere. I think the marginal value of an online version, therefore, would be greater than Ann Arbor, as it would be the only not-in-person option going.

Yes, agreed, but I definitely think they should be on cable.

Although I'm not sure of the exact wording of the law, I was under the impression that cable companies were required to give access to local municipalities free of charge- thus, if we can get video shot and edited into a viewable format, I don't understand why we wouldn't then take it to the most readily digested broadcast medium, first and foremost.

Ann Arbor, by my count, seems to actually have a couple stations- most of which are dead air for the majority of the day- AND they have a studio. Worth contemplating, anyway.

In the meantime, not to sound totally lame, but I'd be happy if the council minutes were just posted online a little quicker, in either plain text or html format.

We have not abandoned the idea of putting meetings on the internet. The project changed when we moved because public meetings held at City Hall are shown live on Channel 12. We have to get a video capture card in order to record the meetings, and plan to store the files at archive.org with the police scanner recordings. The value in having a transcription or recording of public meetings on the internet is in the historical record created. Only the most recent community meetings can be shown on cable.

I heard that in Hamtramck, there used to be public access shows and telethons in addition to meetings on Channel 12, but I believe that ended when positions were cut by the EFM. The City Council asked Comcast to send a representative to one of their meetings last year, but Comcast declined because they only have 1700 customers in Hamtramck. Most people here have satellite dishes for TV.