Libraries needed:
library(rtweet)
library(stringr)
library(dplyr)
library(tidyverse)
library(lubridate)
library(scales)
library(tidytext)tmls <- get_timelines('BernieSanders', n = 3200)
tweets <- tmls %>% select(status_id, source, text, created_at, favorite_count, retweet_count)
tweets <- mutate(tweets,source =ifelse(str_detect(tweets$source,'(?<=Twitter )[^.]*')==TRUE,str_match(tweets$source, '(?<=Twitter )[^.]*'), tweets$source))## # A tibble: 7 x 2
## source n
## <chr> <int>
## 1 iPhone 465
## 2 MediaStudio 377
## 3 Periscope 115
## 4 TweetDeck 465
## 5 VITAppforiOS 2
## 6 WebApp 283
## 7 WebClient 1492
Web Client is the most used source to make tweets, Tweet Deck coming in second and iPhone coming in third place.
Usual time of the day where Twitter source is made from:
According to users, these are BernieSanders favorites tweets:
#favorites
print('Most favorited Tweet:')## [1] "Most favorited Tweet:"
tweets[which.max(tweets$favorite_count), ]$text ## [1] "Donald Trump believes climate change is a hoax. Donald Trump is an idiot."
print('Least favorited Tweet:')## [1] "Least favorited Tweet:"
tweets[which.min(tweets$favorite_count), ]$text## [1] "Yes. And let’s start by getting @senatemajldr to schedule a vote on S.27 - American Miners Act of 2019. We need to keep our promises. @MineWorkers earned these pensions and healthcare. \nhttps://t.co/98W2UoFg0H https://t.co/HFh1wE9FMi"
fav_tweets <- tweets %>% filter(favorite_count > mean(tweets$favorite_count))
count(fav_tweets)## # A tibble: 1 x 1
## n
## <int>
## 1 846
#retweet
print('Most retweeted Tweet:')## [1] "Most retweeted Tweet:"
tweets[which.max(tweets$retweet_count), ]$text## [1] "Donald Trump believes climate change is a hoax. Donald Trump is an idiot."
print('Least retweeted Tweet:')## [1] "Least retweeted Tweet:"
tweets[which.min(tweets$retweet_count), ]$text## [1] "Finishing off the day in Rutland!Thank you @BernieSanders for bringing us all together today, especially in light of the news out of DC. #vtpoli"
re_tweets <- tweets %>% filter(retweet_count > mean(tweets$retweet_count))
count(re_tweets)## # A tibble: 1 x 1
## n
## <int>
## 1 807
## Using `to_lower = TRUE` with `token = 'tweets'` may not preserve URLs.
## # A tibble: 5 x 2
## word n
## <chr> <int>
## 1 #medicareforall 68
## 2 #bernie2020 44
## 3 #demdebate 42
## 4 #nomiddleground 37
## 5 #organizewithbernie 36
The top 20 words used in the 3200 tweets made by BernieSanders with each
frequency
What about the top favorite tweets?
## # A tibble: 7 x 2
## source total
## <chr> <int>
## 1 iPhone 6343
## 2 MediaStudio 4504
## 3 Periscope 1291
## 4 TweetDeck 6313
## 5 VITAppforiOS 23
## 6 WebApp 4027
## 7 WebClient 21409
## Joining, by = "source"
## # A tibble: 13,057 x 6
## source word n tf idf tf_idf
## <chr> <chr> <int> <dbl> <dbl> <dbl>
## 1 VITAppforiOS #demdebate2 1 0.0435 0.560 0.0243
## 2 VITAppforiOS don 1 0.0435 0.336 0.0146
## 3 VITAppforiOS donation 1 0.0435 0.336 0.0146
## 4 VITAppforiOS funding 1 0.0435 0.336 0.0146
## 5 VITAppforiOS guts 1 0.0435 0.336 0.0146
## 6 iPhone @berniesanders 131 0.0207 0.560 0.0116
## 7 Periscope join 66 0.0511 0.154 0.00788
## 8 TweetDeck @berniesanders 81 0.0128 0.560 0.00718
## 9 WebApp @berniesanders 51 0.0127 0.560 0.00709
## 10 VITAppforiOS complex 1 0.0435 0.154 0.00670
## # ... with 13,047 more rows
8 emotions (anger, fear, anticipation, trust, surprise, sadness, joy, and disgust) 2 sentiments (negative and positive) emolex downloaded from http://saifmohammad.com/Lexicons/ We are looking at the percentage of tweets where each sentiment is shown during the hour of the day. A positive sentiment is the highest percentage of tweets
## Parsed with column specification:
## cols(
## X1 = col_character(),
## X2 = col_character(),
## X3 = col_double()
## )
## # A tibble: 3,101 x 3
## status_id source total_words
## <chr> <chr> <int>
## 1 1039662020990324736 WebClient 21409
## 2 1039668059735244800 WebClient 21409
## 3 1039940620746477570 WebClient 21409
## 4 1039986970468990981 WebClient 21409
## 5 1039994004342034438 WebClient 21409
## 6 1040018040048545810 WebClient 21409
## 7 1040029440091271170 WebClient 21409
## 8 1040040865782280192 WebClient 21409
## 9 1040046073962786821 WebClient 21409
## 10 1040226792605401090 WebClient 21409
## # ... with 3,091 more rows
## # A tibble: 6 x 4
## source emotion total_words words
## <chr> <chr> <int> <dbl>
## 1 iPhone anger 6343 338
## 2 iPhone anticipation 6343 371
## 3 iPhone disgust 6343 163
## 4 iPhone fear 6343 391
## 5 iPhone joy 6343 301
## 6 iPhone negative 6343 584
Compared to favorites, no huge difference except tweets from Periscope source are less liked.
## # A tibble: 3,101 x 3
## status_id source total_words
## <chr> <chr> <int>
## 1 1039668059735244800 WebClient 5174
## 2 1042162683825541127 WebClient 5174
## 3 1043611807695421442 WebClient 5174
## 4 1046939970747940864 WebClient 5174
## 5 1047133156921479173 WebClient 5174
## 6 1048032131769929728 WebClient 5174
## 7 1048388524100276224 WebClient 5174
## 8 1048698035172196353 WebClient 5174
## 9 1049803479273811968 WebClient 5174
## 10 1051593847657119745 WebClient 5174
## # ... with 3,091 more rows
## # A tibble: 6 x 4
## source emotion total_words words
## <chr> <chr> <int> <dbl>
## 1 iPhone anger 6343 122
## 2 iPhone anticipation 6343 103
## 3 iPhone disgust 6343 57
## 4 iPhone fear 6343 150
## 5 iPhone joy 6343 91
## 6 iPhone negative 6343 224
