wesslen

import glob
import json
from pymongo import MongoClient

# fill in hostname and port
HOST = "hostname"
PORT = 27017

client = MongoClient(HOST, PORT)
# fill in dbname and colname

wesslen

# install tidyverse if you don't have it
# install.packages("tidyverse")
library(tidyverse)

## Read the csv from a URL
url <- "http://assets.datacamp.com/course/compfin/sbuxPrices.csv"
df <- read_csv(url)

## lubridate package to format the date
# if you get an error below, are you sure you have lubridate? 

wesslen

library(tidyverse); library(lubridate)

url <- "http://nodeassets.nbcnews.com/russian-twitter-trolls/tweets.csv"
tweets <- read_csv(url)

user.url <- "http://nodeassets.nbcnews.com/russian-twitter-trolls/users.csv"
users <- read_csv(user.url)

tweets %>%
  count(Date = as.Date(created_str)) %>%

wesslen

---
title: "Analyzing Russian Trolls: Tidyverse & Text"
author: "Ryan Wesslen"
date: "2/21/2018"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, warning = FALSE)
```

wesslen

labels <- sageLabels(ctmFit, n = 5)
topicsNames <- sapply(1:30, function(x) paste0(labels$marginal$frex[x,], collapse = " + "))

wesslen

## issues: 
## -only gets replies within last ~7 days to the post due to public REST API limits
## -counts don't necessarily align with total replies via browser, perhaps due to private accounts (?)

get_replies <- function(tweetid){
  # get status information for given tweet
  t <- rtweet::lookup_statuses(statuses = tweetid, token = ryan_rtweets)
  
  # use search API to find all tweets directed to the poster
  # and keep only replied to that status

wesslen

library(quanteda); library(ggrepel); library(tidyverse)

ggplotWordcloud <- function(df, maxWords = 50){
  
  corpus(df$text) %>%
    dfm(remove_punct = TRUE, remove = stopwords("English")) %>%
    topfeatures(n = maxWords) %>%
    as.tibble() %>%
    rownames_to_column(var = "word") %>%
    slice(1:maxWords) %>%

wesslen

# Project 1: Charlotte Protest
library(tidyverse); library(lubridate); library(xts); library(dygraphs); library(quanteda)

# 10% sample
protestData <- readRDS("../Protest.RData") %>%
  mutate(time = paste0(substr(postedTime, 1, 13), "00:00 EDT")) %>% # convert to hourly
  mutate(time = ymd_hms(time)) %>%
  select(time, verb, postedTime, body)

# get daily counts

wesslen

library(rtweet)
library(dplyr)

files <- list.files()
files <- files[grep(".json", files)]

getPoints <- function(file){
  parse_stream(file) %>% 
    lat_lng("bbox_coords") %>% # keep bounding box coords
    filter(is_retweet == FALSE & !is.na(lat)) %>% # keep posts and point lat/longs

wesslen

library(tidyverse); library(rtweet)

get_timeline(c("senthomtillis","thomtillis"), n = 3200)  %>%
  mutate(keyword = grepl("bipartisan", text, ignore.case = TRUE)) %>%
  filter(keyword) %>%
  group_by(screen_name) %>%
  ts_plot("3 months", trim = 1L) +
  labs(x= "Date", y= NULL, title = "Thom Tillis' Tweets mentioning `bipartisan*` by Twitter account") +
  theme(text = element_text(size = 12),
        legend.position = c(0.3,0.5))