Monday, November 4, 2013

Using R to find Obama's most frequent twitter hashtags

I've been exploring Jeff Gentry's twitteR package, which has a ton of great functionality for intereacting with twitter data in R. Today, I thought a bit about a problem I've noticed several times on twitter: users profiles are often only noisy signals of the content they tweet about!

I decided that a table of a user's commonly-used tweets might give a better sense of the content a user tweets about. My code to extract the hashtags is below (note: you'll need to load the twitteR package, and complete the OAuth Authentication first.. if you're having trouble with this, try visiting this page)

Here's the code I used:

tw = userTimeline("BarackObama", cainfo = x1, n = 3200)
tw = twListToDF(tw)
vec1 = tw$text
extract.hashes = function(vec){
hash.pattern = "#[[:alpha:]]+"
have.hash = grep(x = vec, pattern = hash.pattern)
hash.matches = gregexpr(pattern = hash.pattern,
                        text = vec[have.hash])
extracted.hash = regmatches(x = vec[have.hash], m = hash.matches)
df = data.frame(table(tolower(unlist(extracted.hash))))
colnames(df) = c("tag","freq")
df = df[order(df$freq,decreasing = TRUE),]
dat = head(extract.hashes(vec1),50)
dat2 = transform(dat,tag = reorder(tag,freq))
p = ggplot(dat2, aes(x = tag, y = freq)) + geom_bar(fill = "blue")
p + coord_flip() + labs(title = "Hashtag frequencies in the tweets of the Obama team (@BarackObama)")

Created by Pretty R at


  1. I like it, but I'd move alpha to alnum for the switch, in case, for example, the hashtag was something like #vote2014

  2. Thanks for the code ! I had an error at first, but adding "stat = "identity" " to the geom_bar() corrected the issue.

  3. Thanks for the code!! i am also getting an error like "Error: stat_count() must not be used with a y aesthetic."