Matthew Ericson at the New York Times did a really cool visualization last week, "The Words They Used", comparing the most frequently used words at the Democratic and Republican Conventions (from the article, "Republicans were more likely to talk about businesses and taxes, while Democrats were more likely to mention jobs or the economy.")

This got me thinking about doing something similar for TV programs. So I did an experiment using the excellent word cloud generator Wordle on transcripts (generated with a single click from a SnapStream TV search appliance for The Daily Show with Jon Stewart and Fox's The O'Reilly Factor with Bill O'Reilly last week (the week of the Republican Convention in Minneapolis). The results:

Monday, September 1, 2008

»The O'Reilly Factor with Bill O'Reilly

The Daily Show with Jon Stewart

(there wasn't a new episode on Monday!)

Tuesday, September 2, 2008

A few notes:

So what do you think? Are these visualizations interesting? What are your observations? I'm not someone who has a background doing content analysis so hopefully I can get some experts to give me their conclusions.

Jeff Jarvis wrote a blog post last week about how he thought the word "actually" was overused on television. Here's the post (he lost a bunch of posts and hasn't restored them all):

Actually is the new ‘y’know’
July 19th, 2008, by Jeff Jarvis

The most overused and unnecessary word on broadcast is “actually.” Start counting how many times it is used by TV people and you’ll hate me for driving you nuts.

While I’m kvetching, why do TV people introduce a panel of three people and then say, “Mr. Jones, let me start with you.” Just start with him: ask your question. Why this need to warn Mr. Jones?

Our TV search appliance is used by a number of journalism schools for content analysis (like Emerson College and GWU, so this got me thinking about how we might try and use our product to measure Jeff's assertion about the word actually.

So here's what I did. I took one week of national TV recordings that we had made on a SnapStream Enterprise TV Server and I did some ad-hoc analysis (remember, I'm not expert on content analysis!) on how frequently the word 'actually' appeared by series and by network.

Here are the 20 shows that use the word 'actually' the most:

And here are the 20 shows that use the word 'actually' the least:

And here's a summary of the the word 'actually' by network:

So at the higher end, offending shows are using the word 'actually' between once every 10 minutes to once every 3 minutes. Since I'm not a content analysis expert, what other trends can you draw? How could the test be improved?

George Washington University’s independent student newspaper, the GW Hatchet, reports on the use of our SnapStream Enterprise product in the School of Media and Public Affairs. The story highlights the impact that SnapStream's television search technology will have at GWU's public policy and journalism schools, making it so that faculty and students can search television broadcasts for educational research and analysis. A few choice quotes:

Sean Aday, an associate professor at GWU, says:

"It's a great tool for research. For example, with the recent news about the U.S. embassy in Serbia, we could collect and compare coverage from all the networks. Graduate students, especially in the research methods class, will be able to conduct their own content analysis.”

Paul Fucito, GWU's director of communications, says:

"It takes seconds and minutes now to do what took weeks or months. After recording, let's say a month of Andersen Cooper, you can then go back type in relevant keywords, find the clips that apply and watch those segments."

Article: SMPA acquires TiVo-like technology