I’ve just returned from a business trip – lots of long days, working sometimes from 8 am to 9 pm or 10 pm. I didn’t get a chance to write this week’s Statistics Sunday post, in part because I wasn’t entirely certain what to write about. But as I started digging into sentiment analysis tools for a fun project I’m working on – and will hopefully post about soon – I found a few things I wanted to share.
The tidytext package in R is great for tokenizing text and running sentiment analysis with 4 dictionaries: Afinn, Bing, Loughran, and NRC. During some web searches for additional tricks to use with these tools, I found another R package: syuzhet, which includes Afinn, Bing, and NRC, as well as Syuzhet, developed in the Nebraska Literacy Lab, and a method to access the powerful Stanford Natural Language Processing and sentiment analysis software, which can predict sentiment through deep learning methods.
I plan to keep using the tidytext package for much of this analysis, but will probably draw upon the syuzhet package for some of the sentiment analysis, especially to use the Stanford methods. And there are still some big changes on the horizon for Deeply Trivial, including more videos and a new look!