issue_classifier/code/prepare_train_merge_tv_fb_gg.R at main · Wesleyan-Media-Project/issue_classifier · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
library(data.table)
library(dplyr)
library(tidyr)
library(stringr)
library(readr)

# Load TV 2022 transcripts with issue codes
path_train <- "../data/tv22_transcripts_issue_codes.csv"
train <- read_csv(path_train)

# Load Meta and Google 2022 hand coded issues
path_fb22 <- "../data/fb_2022.rdata"
load(path_fb22)

path_gg22 <- "../data/gg_2022.rdata"
load(path_gg22)

# Keep only top 25 issues from TV ads
# Specify the columns to keep
columns_to_keep <- c(
  "ISSUE215", "ISSUE10", "ISSUE30", "ISSUE40", "ISSUE212", "ISSUE12", "ISSUE16",
  "ISSUE209", "ISSUE18", "ISSUE83", "ISSUE200", "ISSUE54", "ISSUE95", "ISSUE62",
  "ISSUE55", "ISSUE91", "ISSUE56", "ISSUE53", "ISSUE90", "ISSUE65", "ISSUE45",
  "ISSUE60", "ISSUE58", "ISSUE22", "ISSUE32", "alt", "transcript"
)

# Subset the dataframes
train2 <- train %>%
  select(all_of(columns_to_keep))

fb22 <- fb22_3 %>%
  select(all_of(columns_to_keep))

gg22 <- gg22_3 %>%
  select(all_of(columns_to_keep))

df <- bind_rows(train2, fb22)

fwrite(df, "../data/issues_tv_fb_gg_22.csv")