Show HN: Tracking voter discussion of the Presidential debates | Hacker News Reader

Show HN: Tracking voter discussion of the Presidential debates

live.votizen.com

44 pointsjdunck29 comments

Votizen is working to remove the influence of money in politics by making political decision-making more peer-to-peer.

We're using Twitter and our API to pull tweets from only registered voters, and showing how the discussion breaks down among the parties. This has never been done before as far as I know, not even by the major media outlets.

Please check it out and let me know what you think. If people are interested, we'll work to improve it for the further debates and campaigns.

Open articleView on HN

It's a fun visualization to have up as a dashboard, I think I'll keep it running during the debates.

I love the design; it's very easy to read, and the simplicity and color make it easy to digest information at a glance.

My constructive criticism:

For the word frequency charts at the top of the screen, I'd weed out common words, and/or try to combine words into meaningful phrases. http://www.wordle.net/ does a good job with this (weeding out common words, not discerning phrases).

The number one thing I think people would be interested in would be some kind of sentiment analysis. Perhaps a knowledgable HN'er could suggest a service you could use for this?

From a UX perspective, I'd like the ability to pause and rewind the twitter streams, or filter them for specific words, phrases, geographic locations, or demographics.

One final nitpick, I would suggest using image meta-tags for sharing on Facebook. I posted the live.votizen.com page to my timeline, and there was no image to accompany the link. Here are the meta-tags you can use for richer sharing: http://davidwalsh.name/facebook-meta-tags

But yes, overall, I love this.

We have weeded out a bunch of stuff and are actually handling n-grams, but apparently not with enough weight since they aren't showing up ever.

What words do you think shouldn't be there?

Looks, Look, Good

Also, you should think about stemming - "Looks" and "Look" should be in an equivalence class.

Perhaps we can bang a Snowball stemmer against it and get pretty good coverage on the whole language.

I'd be interested in the trending between points. For example I would just love to see how the trend went after a candidate said certain things and indeed trend based upon the last N minutes. The one which Romney says "I'm gonna (sounded like that too me) stop the subsidy to PBS", Now that would be interesting too me, along with many other statements by them both.

Fascinating stuff, I also noted a few posts positive to romney that wer associated as obama as obama was mentioned and vice a versa. I suspect a first mentioned corralation, though have seen exceptions to that, i'd paste examples but no pause button and late. To truely seperate those type of posts is something were you will need to crowdsource down the line if only to learn a rule set. Though tweets are short so permutations will not be that dynamic as apposed to more open social means and with that the limitation can only help.

Anybody aware of a ruleset that can weight one person over another based upon the content of block of text that mentions both people. Certainly would have uses beyond this. Amazing how something so simple as that involves alot more than you would initialy think.

What you're talking about in the last paragraph is sentiment analysis. This is really cool stuff, but it's tricky. Even if you have 2 humans do sentiment analysis independently, they agree with each other only 70% of the time. Computers tend to do worse, of course.

What makes it extra challenging is that tweets are short, so you have wide error bars on the sorts of math you'd depend on in larger corpuses. I think you'd have to actually bang a parser against it and try to understand what the tweeter is saying.

Btw, here's an example of somebody trying to do sentiment analysis on tweets: http://www.sentiment140.com/search?hl=en&query=romney.

Thank you, sentiment analysis - I just love that term.

As a non-US guy, this is pretty cool.

It's working really well for live updating.

How are you matching Twitter accounts to voter registration?

We have voter rolls (preprocessed) into elastic search. We do search using data available from twitter (name & location, mostly). We can't match all twitter users who are voters, but we have a very low false-positive rate.

And getting lower tomorrow. And then going up in efficacy, while maintaining confidence. There's a ton more we can do with this. :-)

Twitter's actually the hardest nut to crack—pretty sparse user info available. Facebook and LinkedIn have a much higher hit rate, but they don't have such easy streams to grab.

What do you like? What don't you like?

I love the whole concept and execution, except I think the tweets should scroll by slightly slower, as sometimes you don't have time to read a tweet before it disappears.

Yeah, it's a tough call. If we slow it down, we'd have to show fewer as the underlying stream is zipping by. We considered pause or pause-on-hover, but sort of assume people will glance at it, not stare at it, if that makes sense. Gestalt, not reading?

Nice! But on a netbook if I reduce the font to try to get both the Democrip and Rebloodlican halves on the screen, the Rebloodlican side just disappears.

Also, filtering options would be great, for example only viewing tweets from independents, or only ones that mention Romney.

Good ideas, all. There are a couple more debates. :-)

In the meantime, if you want to play with filters, feel free to try this super-alpha toy I threw together the other day: https://www.votizen.com/twitter-fight/. It works on a rolling 24-hour window of Romney- or Obama-mentioning tweets. Buggy as heck, but fun to play with!

I was wondering what the time window is for the "voters are saying" bar charts. Seemed too long and static... wish those charts were more dynamic.

It is cumulative. I think I agree with you, though. I'll think about how to do rolling for the next debate (Oct 11.)

This is clever and I find it highly entertaining to read all of the trolls and flamebait. The white flash when the graphs update is extremely irritating to me.

Fellow Votizen dev.

I agree about the white flash. We decided to give nvd3 a try for charts, and it does that as part of its default animation. We'll probably disable animations or figure out some way to keep redraws from flashing like that.

Add this to the CSS:

div.chart {opacity: 1 !important;}

It is a hack but should work.

Whoops, no, I did the flash to give a pulse impression. The transition is unavoidable, but I did the opacity animation on purpose. :)

I removed the on-update flash.

I like the idea and the execution, I hate reading the quality of the tweets. I try not to be a snob, but yikes.

Interesting. I'm particularly interested in how you got this matched up against voter lists. Some states are quite a PIA to collect.

You're telling me. :-) We spent 2 years getting all those lists together and normalizing them. We're midway through a refresh now—new normalization framework, new matching algorithms, etc., etc.

Reply on news.ycombinator.com