So I know I was going to stick just to Text Analysis and not talk about visualization, but I saw a really cool StarTree use (hyperbolic tree) yesterday. The team has automated the process of putting the whole BusinessObjects org chart into a StarTree, including photos of employees, etc. This makes it really easy to see who has too many direct reports, who has no direct reports, and who works for who within the organization. Kudos to the team! For more information on StarTrees, visit http://www.businessobjects.com/product/catalog/startree/ (you can also find a few demos at http://www.inxight.com/products/sdks/st/.)
Back to Text Analysis... One of the applications that's up and coming for us is Voice of the Customer. So I thought I'd take a lighthearted look at the voice of the customer on American Idol. A quick Text Analysis run of the "top 12 women on Idol" blogs reveals that the majority of posters think Alexandrea and Amanda should be going home tonight, with Kristy Lee Cook a distant third. Now my own theory is that being noticed at all is far better on the show than being forgotten, since you vote for who you like as opposed to against those you don't like. Which means that it's more likely Kady, Syesha, and Alaina would be the three most likely to go home, since they are mentioned the least.
One of the cool things about Text Analysis is that you don't need to feed it lists of terms for it to track them. For example, in spite of the fact that some of the contestants have the STRANGEST names (Asia'h? Syesha?), only 1 of the 12 girls is not consistently automatically detected by the system as being a PERSON -- Ramiele Malubay. She's identified as a "miscellaneous noun group", which makes it easy to locate her and then add her to a name catalog, ensuring that in subsequent runs, I will find her as a person.
(By the way, according to the "being remembered is best" rule, should be in the top 2 at the end of the competition unless she really messes up. She's mentioned the most of all the girls).
Some other interesting things that Text Analysis dug up...
- There is one slightly inappropriate obscene phrase on the blogs that the moderator didn't catch.
- The hottest sentiments and topics of the last show were "song choice", "very sweet", "soooo bad" (with and without the trailing oooo's).
- The worries about some singers having a professional background seem to have trailed off some compared to last week
Anyway, this just goes to show that Text Analysis doesn't have to be all work and no play.
One fun thing everyone can try out is to see what's being discussed in their own emails and items on their desktop. Simply visit http://labs.businessobjects.com/ and try out the "Intelligent Search for Google Desktop" application.
