Data Mining: Text Mining, Visualization and Social Media

YouTube by the Numbers

Tim Wintle of Rubberductions forwarded me a pointer to a new piece of research which analyses viewership for YouTube videos in the first month. Key findings:

In the first month on YouTube

• 70% of videos get at least 20 views
• 50% of videos get at least 100 views
• Fewer than 20% of videos get more than 500 views
• Fewer than 10% of videos get more than 1, 500 views
• 3% of videos get more than 25, 000 views
• Around 1% of videos get more than 500, 000 views

These numbers are described by the following chart:

Race and Gender in the Election

A quick look at BlogPulse paints an interesting picture regarding the focus of attention on race and gender in the election (thanks to Jason Adams for improving the queries!).

Note that variations on these queries (e.g. "first female US president") don't yield any significant extra volume. Note also how the above differs from the trends for the names of the candidates.

 

Wired - The Secret Life of a Blog Post

The current edition of Wired has an article which Frank Rose and I wrote describing what happens to a blog post once you hit the 'publish' button. The article has now made it to the web in the form of a snappy flash graphic which allows you to move around the diagram to focus on the details.

I've been chatting off and on with Frank for a while about the social media space and it's great to see something make it to the magazine.

TechMeme - Beyond the List

While there may be plenty to say about the tyranny of preferential attachment, TechMeme has provided a compelling application that presents data in something other than the simple list. Recently, this video appeared on TechMeme's news blog. It shows a time lapse of the state of TechMeme - stories bubble up to the top of the site, and the conversations around those stories build up around them, consuming more and more real estate.

Bush Versus The Election

Using a query which captures all the candidates intersected with terms that indicate the presidential election and comparing the trend with that for the term 'bush' we get some idea of what a lame duck president has to deal with. No wonder Nixon Bush is off running around outside his cage trying to get some attention (not to mention relevance).

Which Party Gets the Buzz?

By forming a query out of the candidate names for each party:

• (clinton OR obama OR edwards OR gravel OR kucinich)
• (giuliani OR huckabee OR keyes OR mccain OR paul OR romney OR thompson)

and combining this with terms that are commonly used around the presidential election:

• (president OR presidential OR campaign OR election OR primary OR primaries)

we can get some understanding of where attention lies at the party level.

 

The graph above suggests that until November both parties were getting a similar amount of attention. Prior to the first batch of primaries, the Republican candidates were getting a little more, but the Democratic race may be proving a more engaging topic than the Republican one.

You can play around with this query to see where attention lies for different issues by adding in terms such as +health, +war, +iraq, +immigration, etc.

ReadBurner Burner

ReadBurner provides a social rank for items shared via link blogs. Now is the right time to mention ReadBurner as it is itself at the top of the list of shared items due to this post on the Google Operating System.

[thanks, Justin]

An Indirect Democracy

I've just come across NPR's interactive primaries election map. Worth checking out - they do a reasonable job with providing a summary/aggregate view, and interactions that drill down to individual results.

Something that I've learned as an expat living in the USA - democracy here isn't about the equal weight of each citizen's voice. The most obvious example was in the previous election for president in which the person with the most votes (Al Gore) lost to the person with the least votes (George Bush). Another strange approach to democracy is evident in the primary elections. These are the elections by which parties determine who their candidate is. These elections are carried out in sequence. This has the effect of the earlier elections (like Iowa, New Hampshire, etc.) impacting the results of the later election - with the electorat being swayed by information determining who is really in the running and who is not.

Tealeaves

Not quite sure how to interpret these, but they seem relevant.

 

Geeking with Greg

Exciting news: Greg Linden is going to be joining us here at Live Labs next week. Congratulations to Greg!