library(tidyverse)
#library(dsl)
library(dplyr)
dsl_csv <-"~/dsl/121625_DSL_frame.csv"
dsl_df <- read.csv(dsl_csv, header = TRUE) 

dsl_df <- dsl_df |>
  filter(isAuthorWMF != "BzImport")

dsl_df_long <- dsl_df %>%
  pivot_longer(
    cols = c(olmo_EP_prop_adac, olmo_RK_prop_adac, olmo_TSOL_prop_adac),
    names_to = "tag",
    values_to = "proportion"
  ) %>%
  mutate(tag = gsub("olmo_|_prop_adac", "", tag),
         tag = case_when(
           tag == "EP" ~ "Existent Problem",
           tag == "RK" ~ "Record Keeping",
           tag =="TSOL" ~ "Solutions"
         )) 

olmo_comparison <- ggplot(
  dsl_df_long,
  aes(
    x = tag,
    y = proportion, 
    fill = isAuthorWMF,
  )
) + 
  facet_grid(source ~ ., 
             scales = "free_y",
             labeller = labeller(source = c("c1" = "VisualEditor", 
                                            "c2" = "HTTPS-login", 
                                            "c3" = "HTTP-deprecation"))) +
  geom_boxplot() +
  theme_minimal() + 
  scale_fill_viridis_d() +
  labs(
    x = "Tag", 
    y = "% of sentences tagged", 
    color = "Is Author WMF?",
    fill = "Is Author WMF?"
  ) + 
  theme(legend.position = "top")
olmo_comparison
ggsave(
  filename = "121625_machine_label_comparison.png",
  plot = olmo_comparison,
  width = 12,    # inches
  height = 6,   # inches
  dpi = 800     # high resolution
)