I stumbled across an interesting book the other day called Tidy Text Mining. It was nice to find something in R that looks intuitive to me.
I've always wondered whether RBA minutes are indicative of future movements in interest rate prices. There is certainly enough time spent attempting to decode RBA releases to make something like this worthwhile. Maybe some of those poor analysts could free up a bit more of their time?
Anyway - I wrote up something simple that ingested the minutes beginning 2011 and finishing 2017 and charted the sentiment of the individual releases over time.
The above shows the trends of counts of positive words, negative words and (Positive - Negative) over the corpus.
Interestingly, the RBA keeps with its dour reputation by always being more negative than positive (except for one lonely month!).
As always, I'd love to hear any comments about this analysis - and the code can be found on my github.