I have been been experimenting with using language analysis tools like IBM’s Watson when I’m writing stuff for clients, and on stuff they write too. I think this kinda thing is going to be more and more useful on the creative side of advertising, not just for copywriting but also for visual design.
I’m learning about it so that I will know how to control it and use it just like any other tool, not just assume it’s an end-all-be-all “answer.”
And really, I figure the more I learn about things like artificial intelligence, machine learning, and predictive analysis the better off I’ll be when the inevitable robot uprising happens. Hopefully. Maybe?
I asked IBM’s Watson what it thought of Elise’s last blog post.
Anyhow, I have been thinking for a while that it would be really neat to write a plugin for our clients’ websites that would process the content they create through Watson (or something like it) as they create it, and kinda give them data about the emotional tone of what they are writing, and how it might be perceived.
Here’s an example; I asked Watson what it thought of Elise’s last blog post.
Watson scored a whole bunch of topics and concepts in the post on a scale from most positive to most negative sentiments. Here’s just a few:
So going by the chart above, her post’s most favorable sentiments are related to wine and celebration, and its least are about politicians and California, lol.
This is where things start to get more interesting. Watson calculated the emotional makeup of each individual sentiment expressed in Elise’s post. Here’s the emotional makeup of the negative vacation sentiment from the above chart:
So cross-referencing these two bits of data tells us that, even though the overall sentiment regarding vacations in Elise’s post was negative, looking deeper reveals that she actually had connected them with the positive emotion joy, with the second most prominent emotion being sadness.
It makes sense given the post’s content; for the last year she has only been able to vacation in Missouri, and only to visit friends getting married or having babies.
This stuff is not completely reliable of course, but it’s still pretty interesting.
The cost of owning cars
As I went further down the Watson rabbit hole, I found a little function that takes a large sample of a person’s writing and stirs it all up in some kinda algorithm of doom and spits out what it believes is the core make up of that person’s personality. Well that’s just creepy as fnck.
I took the last couple years of my posts here and shoved them in Watson’s face.
Naturally I had to try it.
I took the last couple years of my posts here and shoved them in Watson’s face. It thought about for a bit and then started to tell me stuff about myself.
- I am philosophical. Yeah I suppose.
- I challenge authority. That works, I guess.
- I am unconcerned with both tradition and achieving success. Not really sure those should both be on the same line item, but Watson thought they should be.
- I am sensitive to ownership cost when buying automobiles. Umm.
I’ve never bought a car.