MODELS <- here("exploratory_studies/13_europarl/gender_ratings/")
all_model_ratings <- map_df(list.files(MODELS, full.names = T), fread)  %>%
  mutate(epochs = map(model, ~str_split(pluck(pluck(str_split(., "en_"),1), 2), "ep")),
         epochs = map(epochs, ~pluck(pluck(.,1),1)),
         model_type = case_when(str_detect(model, "trans_f") ~ "translated",
                                str_detect(model, "untrans") ~ "untranslated")) %>%
  select(model_type, epochs, word, male_score) %>%
  mutate(epochs = as.character(epochs))

GENDER_NORMS <- here("data/study1a/raw/GlasgowNorms.csv")
glasgow_norms <- read_csv(GENDER_NORMS) %>%
  select(word, GEND_M) %>%
  rename(human_gender_rating = GEND_M) %>%
  rowwise() %>%
  mutate(word =  str_split(word, " ", simplify = T)[1],
         word = tolower(word)) %>%
  distinct() %>%
  ungroup()

all_ratings <- all_model_ratings %>%
  left_join(glasgow_norms) 

ggplot(all_ratings, aes(x  = human_gender_rating, y = male_score)) +
  geom_point(size = .2, alpha = .2) +
  geom_smooth(method = "lm") +
  facet_grid(epochs ~ model_type) +
  theme_bw()

cor_values <- all_ratings %>%
  mutate(epochs = as.numeric(epochs)) %>%
  group_by(model_type, epochs) %>%
  nest() %>%
  mutate(temp = map(data, ~tidy(cor.test(.$human_gender_rating, .$male_score)))) %>%
  select(-data) %>%
  unnest() 


kable(cor_values)

model_type	epochs	estimate	statistic	p.value	parameter	conf.low	conf.high	method	alternative
translated	10	0.1337143	8.903065	0.0e+00	4354	0.1044310	0.1627658	Pearson’s product-moment correlation	two.sided
translated	20	0.1117646	7.421269	0.0e+00	4354	0.0823400	0.1409946	Pearson’s product-moment correlation	two.sided
translated	5	0.1340506	8.925870	0.0e+00	4354	0.1047698	0.1630992	Pearson’s product-moment correlation	two.sided
untranslated	10	0.0802058	4.949089	8.0e-07	3783	0.0484700	0.1117798	Pearson’s product-moment correlation	two.sided
untranslated	20	0.0726241	4.478651	7.7e-06	3783	0.0408590	0.1042425	Pearson’s product-moment correlation	two.sided
untranslated	5	0.1023304	6.327158	0.0e+00	3783	0.0707013	0.1337539	Pearson’s product-moment correlation	two.sided

ggplot(cor_values, aes(y = estimate, x = epochs, group = model_type)) +
  facet_wrap(~model_type) +
  geom_point() +
  geom_line() +
  theme_bw()