Twitter by Location

Make functions that will gather tweets by keyword in multiple geographic locations, find the most resonating and most activating tweets in each city and present them in a pretty table for reading

Jeremy Allen https://jeremydata.com
05-14-2020


Problem: What are Twitter users saying about COVID-19 in different cities across the US?


Load our packages. rtweet retrieves our tweets from the Twitter API and also gets the latitude and longitude coordinates for our cities by sending our city names through the Google Maps API. We will use the reactable package to make beautiful custom tables to read our top tweets in each city.


# do we need to get more tweets?
need_online_tweets <- FALSE
if(need_online_tweets == FALSE) {
  tweets <- readRDS(fs::dir_ls(here::here("_posts",
                               "2020-05-14-twitter-by-location"),
                               regexp = ".*covid_tweets.*"))
} else { # download more tweets:

#---- some helper functions for later
  
# search twitter
get_tweets <- function(locs, query) {
  
  search_tweets(
    q = query,
    geocode = locs,
    n = 2000,
    include_rts = FALSE
  )
  
}

# get coordinates for the places we want to search
get_us_coords <- function(place, mile_radius) {
  
  # function output should be a character vetcor 
  # that looks like: "40.397408,-102.054770,130mi"
  # with no spaces, which is the format required by
  # the twitter api
  
  if(!inherits(place, "character")) {
    stop ("place must be a quoted character string")
  }
  
  if(!inherits(mile_radius, "character")) {
    stop ("mile_radius must be a quoted character string")
  }

  # requires google maps api key
  rtweet::lookup_coords(
    address = place,
    components = "country:US",
    apikey = Sys.getenv("GOOGLE_MAPS_KEY")
    ) %>%
    .[["point"]] %>%
    paste0(., collapse = ",") %>%
    paste0(., ",", mile_radius, "mi")
  
}

#--- now to using those functions to get the tweets we want

# list of places to search for tweets
my_places <- list(
  "New York, New York",
  "Atlanta, Georgia",
  "Chicago, Illinois",
  "Dallas, Texas",
  "Denver, Colorado",
  "Phoenix, Arizona",
  "Seattle, Washington",
  "San Francisco, California"
)

# map our get_us_coords() function over our list of places to get a list of
# lat long points with bundled radius for each place
locs <- purrr::map(
  my_places,
  get_us_coords,
  mile_radius = "55"
)

time_of_search <- Sys.time()

# map our get_tweets() function over our list of places
# then combine results and remove duplicate tweets
dat_list <- purrr::map(
  .x = locs,
  .f = get_tweets,
  query = "COVID"
) %>% # name each list element with its location name, to use in bind_rows below
  set_names(my_places)

dat <- bind_rows(
  dat_list,
  .id = "city_searched" # this column will be filled with the list element names
) %>% unique()


tweets <- dat %>% 
  select( # reorder columns
    city_searched,
    location,
    created_at,
    screen_name,
    name,
    text,
    hashtags,
    favorite_count,
    retweet_count,
    quote_count,
    reply_count,
    status_id,
    everything()
  ) %>% # the geo_coords column is a list column of lat-lng vectors, so unnest_wider
  mutate(geo_coords = map(geo_coords, ~set_names(., c("lat", "lng")))) %>% 
  unnest_wider(geo_coords) %>% 
  arrange(
    desc(created_at)
  )

#---- write the tweets to disk

# format time for use in file name
my_time <- str_replace_all(time_of_search, " ", "_")
my_time <- str_replace_all(my_time, ":", ".")
my_file_name <- paste0("covid_tweets_", my_time)

# write tweets to disk
saveRDS(tweets, here::here(my_file_name))
#rio::export(tweets, here::here(my_file_name))

}


Group by city, arrange the favorite_count column in descending order, and limit the columns we display.


# favorited
most_resonating <- tweets %>% 
  select(city_searched,
         text,
         favorite_count,
         hashtags,
         screen_name,
         created_at,
         retweet_count) %>% 
  group_by(city_searched) %>%
  arrange(desc(favorite_count)) %>% 
  slice(1:10)
  

# retweeted
most_activating <- tweets %>% 
  select(city_searched,
         text,
         retweet_count,
         hashtags,
         screen_name,
         created_at,
         favorite_count
         ) %>%
  group_by(city_searched) %>%
  arrange(desc(retweet_count)) %>% 
  slice(1:10)


Set some basic CSS for our table class and the class of our text column.


.tweet-tbl {
  font-size: .6em;
}

.text-col {
  font-weight: 600;
  color: #e96384;
}


The table showing the top 10 most favorited tweets in each city. Click a city to expand its rows. Scroll right to see more columns. First up are the tweets that resonate the most, i.e., the tweets most favorited.


# a color palette function to shade table cells based on a value
my_pal <- function(x) rgb(colorRamp(c("#f0f0f5", "#a3a3c2"))(x), maxColorValue = 250)

resonate_table <- reactable(
  most_resonating,
  groupBy = "city_searched",
  filterable = TRUE,
  class = "tweet-tbl",
  height = 600,
  columns = list(
    text = colDef(
      class = "text-col", # see CSS chunk for styling
      minWidth = 400
    ),
    favorite_count = colDef(style = function(value) {
      normalized <- (value - min(most_resonating$favorite_count)) / 
        (max(most_resonating$favorite_count) - min(most_resonating$favorite_count))
      color <- my_pal(normalized)
      list(background = color)
      }
    ),
    hashtags = colDef(style = list(borderLeft = "2px solid #8585ad"))
  )
)

Top 10 most resonating tweets in each city


Next up are the tweets that activate the most, i.e., the tweets most retweeted.


activate_table <- reactable(
  most_activating,
  groupBy = "city_searched",
  filterable = TRUE,
  class = "tweet-tbl",
  height = 600,
  columns = list(
    text = colDef(
      class = "text-col", # see CSS chunk for styling
      minWidth = 400
    ),
    retweet_count = colDef(style = function(value) {
      normalized <- (value - min(most_activating$retweet_count)) / 
        (max(most_activating$retweet_count) - min(most_activating$retweet_count))
      color <- my_pal(normalized)
      list(background = color)
      }
    ),
    hashtags = colDef(style = list(borderLeft = "2px solid #8585ad"))
  )
)

Top 10 most activating tweets in each city

Reuse

Text and figures are licensed under Creative Commons Attribution CC BY 4.0. The figures that have been reused from other sources don't fall under this license and can be recognized by a note in their caption: "Figure from ...".

Citation

For attribution, please cite this work as

Allen (2020, May 14). jeremydata: Twitter by Location. Retrieved from https://jeremydata.com/posts/2020-05-14-twitter-by-location/

BibTeX citation

@misc{allen2020twitter,
  author = {Allen, Jeremy},
  title = {jeremydata: Twitter by Location},
  url = {https://jeremydata.com/posts/2020-05-14-twitter-by-location/},
  year = {2020}
}