I just saw Sonal of a16z tweet out her favorite podcast, so I had to take a look.
It was about a particularly interesting application of machine learning—one that’s fascinating to me on numerous levels—the ability to predict the success of a written campaign by evaluating its style, vocabulary, and typography.
The research they spoke about looked at lots of different types of writing, but the one I thought was interesting was Kickstarter campaigns. They were able to predict with 90% effectiveness which projects would become successful.
They also looked at job listings, and found all sorts of things there as well.
From all of this they came away with terms and phrases that are good and bad to use in each medium, and the researcher said her team had been approached by some Hollywood types to get them to analyze screenplays.
Think about looking at all screenplays, going backward, with the knowledge of which ones became successful. As they point out, there is noise that can enter into such analysis. Maybe the movie got lucky and it wasn’t the writing, for example.
Anyway, the point is that there are so many potential applications of not only this type of writing analysis, but of machine learning in general.
I agree with Sonal; it was a phenomenal show. I’d love to see more shows on interesting, practical applications of data science and machine learning.
- Here’s the link to the episode.