Hacker News new | ask | show | jobs
Our startup analyzed the debate last week. What do you think? (heartbyte.com)
40 points by boorad 5252 days ago
14 comments

After last week's debate I looked for an app where I could get objective, near-real time analysis pertaining to the accuracy of a candidate's claim, since Wolf Blitzer obviously wasn't interested in doing his job. To my surprise, I couldn't find an app that did what I wanted. I considered writing one, but it's been almost a week and I haven't done anything yet, so I'll share the idea and maybe someone else will build it.

The basic idea would be that you press a button whenever a candidate says something you think is controversial. This will bring up a list of recent controversial statements made by the candidates, and you have the option to select one. Meanwhile, someone on the server side is using his best google-fu to immediately find primary sources that are relevant to each controversial statement. When a primary source is found, it is immediately pushed out to the user.

A lot of debate points really aren't "true" or "false" but only "true from a certain point of view". There's exceptions in both directions, certainly, but I don't think this is amenable to real-time analysis by anyone; you're mostly going to get fed the biases of the person on the other end of the wire, not "truth".
Interesting technology. But my biggest complaint is the font. I absolutely hated the font of the blog post, and it made it almost unreadable for me. Maybe I'm too old school, but the weird T's and the too-small-and-narrow fonts made my head hurt. I forced myself to read the entire blog post so that I could have legitimacy in writing this comment, but otherwise I would have just closed my window.
Really don't see why steve8918's comment was downvoted.

On a Win 7 machine, running Chrome 16.0.912.75, the font would appear to be a legitimate problem. Were this a page that I had put together, I'd want to know about it.

uncle. fixed.
Feature suggestion: spectral analysis of the debaters' voices: http://www.www.asanet.org/images/members/docs/pdf/featured/0...
nice. is there any more info about your company? http://heartbyte.com is kinda minimal (even with the comments) and http://heartbyte.com/blog gives an error.

if not, can you outline what technology you're using (i see you've commented on couchdb, some time ago...)? how quickly can you provide results? what does the user interface look like from either side? (i'm a technical guy, but i bet others here would also like to know how you're funded etc...)

We have some rough edges on our web presence to be sure. This may help a bit on the UI front: http://heartbyte.com/s/hn

As for the tech, given the requests we need to handle from our distributed queue, we do clustered stream processing in near realtime. Our broadcast customers want a line on their screen within a second or two.

And for now we are bootstrapped.

Just curious, but your age ranges are 8 years, 17 years, 14 years, and Unknown. How did you come by those buckets? I can't figure out the intended distinction between each group.
Just an FYI, all your results are in percentages, where the total is never given throughout the article. 'about 30' participants is a throwaway gesture - if one wants to actually gauge the quality of the stats, you need an accurate total, and surely 'about 30' could be replaced with the single figure?

Otherwise, looks neat!

Nice to see that someone is actually applying technology to solve real problems! Great work.
Excellent work, guys. This is going to be AWESOME for the upcoming debates!
Interesting choice of color on the age graph. It almost looks like it implies those older than 41 are not somehow as important those younger. They almost blend into the background.
Minor nit: the distribution of Republican, Democrat, and Independent in the population at large isn't 1/3rd each, so your sample is a bit off from representative.
Cool stuff. I can see this being a fun interactive way to participate in tons of reality shows...e.g. Project Runway.
Seems like it has potential, but using an example culled from 30 non-random data points isn't particular interesting.
feed subscription is broken.