CBFC Watch Summaries

# This is pre-processing you can ignore
library(tidyverse)

## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr     1.1.4     ✔ readr     2.1.5
## ✔ forcats   1.0.0     ✔ stringr   1.5.1
## ✔ ggplot2   3.5.2     ✔ tibble    3.3.0
## ✔ lubridate 1.9.4     ✔ tidyr     1.3.1
## ✔ purrr     1.1.0     
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag()    masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors

library(knitr)
library(rsample)
data <- read_csv("https://github.com/diagram-chasing/censor-board-cuts/raw/refs/heads/master/data/data.csv", 
                 col_types = cols(.default = "c")) %>%
  mutate(
    cert_date = as.Date(cert_date),
    total_modified_time_secs = as.numeric(total_modified_time_secs),
    deleted_secs = as.numeric(deleted_secs),
    replaced_secs = as.numeric(replaced_secs),
    inserted_secs = as.numeric(inserted_secs)
  ) %>%
  mutate(
    office = str_split(certifier, ",") %>%
      map_chr(last) %>%
      str_trim()
  ) %>%
  separate_rows(ai_content_types, sep = "\\|") %>%
  mutate(ai_content_types = str_trim(ai_content_types))%>%
  mutate(
    language = case_when(
      language == "Oriya" ~ "Odia",
      language == "Gujrati" ~ "Gujarati",
      language == "Chhatisgarhi" ~ "Chhattisgarhi",
      language == "Hariyanvi" ~ "Haryanvi",
      language == "Hindi Dub" ~ "Hindi Dubbed",
      TRUE ~ language
    )
  ) %>%
  filter(!is.na(language))

top_languages <- data %>%
  filter(!is.na(language), language != "") %>%
  distinct(certificate_id, language) %>%
  count(language) %>%
  slice_max(order_by = n, n = 15) %>%
  pull(language)

1 Percentage of action types

data %>%
  filter(!is.na(ai_action)) %>%
  count(ai_action, name = "count", sort = TRUE) %>%
  mutate(`Percentage (%)` = count / sum(count) * 100) %>%
  rename(`Action Type` = ai_action, Count = count) %>%
  kable(digits = 1)

Action Type	Count	Percentage (%)
deletion	45443	43.7
audio_modification	22247	21.4
insertion	15071	14.5
visual_modification	8326	8.0
replacement	6609	6.4
text_modification	5740	5.5
content_overlay	439	0.4

2 Percentage of content altered by type

data %>%
  filter(!is.na(ai_media_element)) %>%
  count(ai_media_element, name = "count", sort = TRUE) %>%
  mutate(`Percentage (%)` = count / sum(count) * 100) %>%
  rename(`Media Type` = ai_media_element, Count = count) %>%
  kable(digits = 1)

Media Type	Count	Percentage (%)
visual_scene	47869	46.1
text_dialogue	35207	33.9
metadata	16034	15.4
music	4211	4.1
other	554	0.5

3 Share of ratings by language

ratings_by_language <- data %>%
  mutate(rating_group = case_when(
    rating %in% c("UA", "UA 7+", "UA 13+", "UA 16+") ~ "General Audience (U/UA)",
    rating == "A"                                         ~ "Adults Only (A)",
    rating == "U"                                         ~ "Unrestricted (U)",
    rating == "S"                                         ~ "S",
    TRUE                                                  ~ "NA / Other"
  )) %>% 
 filter(language %in% top_languages) %>%
  distinct(certificate_id, language, rating_group) %>%
  count(language, rating_group) %>%
  group_by(language) %>%
  mutate(percentage = n / sum(n)) %>%
  ungroup() %>%
  select(-n) %>%
  pivot_wider(names_from = rating_group, values_from = percentage, values_fill = 0) %>% 
  mutate(across(where(is.numeric), ~ .x * 100))

ratings_by_language%>%
  kable(digits = 1)

language	Adults Only (A)	General Audience (U/UA)	Unrestricted (U)	NA / Other	S
Bengali	7.2	74.7	18.1	0.0	0
Bhojpuri	9.0	85.1	5.9	0.0	0
Chhattisgarhi	3.7	72.0	24.4	0.0	0
English	16.1	72.3	11.7	0.0	0
Gujarati	2.2	71.9	26.0	0.0	0
Hindi	9.5	78.1	12.2	0.3	0
Hindustani	0.9	93.0	6.0	0.0	0
Kannada	11.2	71.3	17.5	0.0	0
Malayalam	4.8	60.1	34.8	0.3	0
Marathi	6.9	73.8	19.3	0.0	0
Odia	1.1	56.8	42.0	0.0	0
Punjabi	5.6	69.4	25.0	0.0	0
Tamil	6.6	67.7	25.4	0.4	0
Telugu	12.4	74.7	12.9	0.0	0
Urdu	1.4	91.3	7.2	0.0	0

4 Duration modified by language

Total time modified by language:

In your SQL code, you grouped it by rating which I have not done; I have just calculated by grouping movies by language. But the values are comparable.

movie_durations <- data %>%
  filter(!is.na(language), language != "") %>%
  distinct(certificate_id, language, duration_secs) %>%
  mutate(duration_secs = as.numeric(duration_secs)) %>%
  group_by(language) %>%
  summarise(total_movie_secs = sum(duration_secs, na.rm = TRUE))

modified_durations <- data %>%
  mutate(total_modified_time_secs = as.numeric(total_modified_time_secs)) %>%
  filter(!is.na(language), language != "") %>%
  group_by(language) %>%
  summarise(total_modified_secs = sum(total_modified_time_secs, na.rm = TRUE))

language_summary <- movie_durations %>%
  full_join(modified_durations, by = "language") %>%
  mutate(
    `Total Movie Duration (Hours)` = total_movie_secs / 3600,
    `Total Modified Duration (Hours)` = total_modified_secs / 3600,
    `Percent Modified (%)` = (total_modified_secs / total_movie_secs) * 100
  ) %>%
  select(
    Language = language,
    `Total Movie Duration (Hours)`,
    `Total Modified Duration (Hours)`,
    `Percent Modified (%)`
  ) %>%
  filter(`Language` %in% top_languages) %>% 
  arrange(desc(`Total Movie Duration (Hours)`))

language_summary %>%
  kable(digits = 1)

Language	Total Movie Duration (Hours)	Total Modified Duration (Hours)	Percent Modified (%)
Hindi	8351.4	192.6	2.3
Tamil	5534.6	63.3	1.1
Telugu	5405.3	55.8	1.0
Kannada	4223.3	52.9	1.3
English	2876.2	63.9	2.2
Malayalam	2852.5	53.2	1.9
Bhojpuri	2358.2	84.1	3.6
Marathi	1520.9	26.3	1.7
Bengali	1033.8	27.9	2.7
Gujarati	775.1	4.9	0.6
Odia	557.9	11.9	2.1
Punjabi	548.4	5.3	1.0
Hindustani	471.8	64.5	13.7
Chhattisgarhi	191.1	0.8	0.4
Urdu	143.4	17.6	12.3

Average time modified by language:

movie_modifications <- data %>%
    filter(!is.na(language), language != "", total_modified_time_secs > 0) %>%
    group_by(certificate_id, language) %>%
    summarize(
        total_duration_modified = sum(total_modified_time_secs, na.rm = TRUE),
        .groups = "drop"
    )
top_languages <- movie_modifications %>%
    count(language) %>%
    slice_max(order_by = n, n = 10) %>%
    pull(language)


movie_modifications %>%
    filter(language %in% top_languages) %>%
    bootstraps(times = 2000) %>%
    mutate(movie_data = map(splits, analysis)) %>%
    unnest(movie_data) %>%
    group_by(language, id) %>%
    summarize(avg_duration = mean(total_duration_modified), .groups = "drop") %>%
    group_by(language) %>%
    summarize(
        mean = mean(avg_duration),
        conf.low = quantile(avg_duration, 0.025),
        conf.high = quantile(avg_duration, 0.975)
    ) %>%
    mutate(across(c(mean, conf.low, conf.high), ~ .x / 60)) %>%
    arrange(desc(mean))%>%
  kable(digits = 1)

language	mean	conf.low	conf.high
Hindustani	18.0	16.9	19.1
Bhojpuri	6.5	5.9	7.1
Bengali	5.4	4.5	6.2
Malayalam	4.0	3.5	4.6
Marathi	3.8	3.1	4.4
Hindi	3.7	3.5	3.9
Kannada	3.4	1.8	6.2
English	2.6	2.5	2.8
Telugu	2.6	2.4	2.9
Tamil	2.3	2.1	2.5

5 Avg modifications per film (by language)

I generally avoid saying cuts because these could just be visual disclaimers being added and so on, not necessarily a deletion/replacement. Also mapped with confidence intervals based on this example: https://github.com/dgrtwo/data-screencasts/blob/master/bird-collisions.Rmd

avg_cuts_by_lang <- data %>%
  filter(!is.na(language), language != "") %>%
  group_by(language) %>%
  summarise(
    total_movies = n_distinct(certificate_id),
    total_cuts = n(),
    .groups = "drop"
  ) %>%
  filter(total_movies > 50) %>%
  mutate(avg_cuts = total_cuts / total_movies) %>%
  arrange(desc(avg_cuts))

avg_cuts_by_lang %>%
  kable(digits = 1)

language	total_movies	total_cuts	avg_cuts
English	1687	16446	9.7
Tamil	2671	16309	6.1
Bhojpuri	1031	6240	6.1
Telugu	2558	14745	5.8
Bengali	498	2714	5.4
Hindustani	215	1078	5.0
Hindi	4178	20915	5.0
Kannada	1975	9886	5.0
Marathi	737	3584	4.9
Urdu	69	325	4.7
Malayalam	1338	5683	4.2
Punjabi	268	1084	4.0
Gujarati	366	1413	3.9
Chhattisgarhi	82	315	3.8
Odia	264	972	3.7

Maybe graphing these with confidence intervals? Which ones have the highest spread?

library(tidyverse)
library(rsample)

set.seed(2025)

movie_cuts <- data %>%
  filter(!is.na(language), language != "") %>%
  count(certificate_id, language, name = "n_cuts")

top_languages <- movie_cuts %>%
  count(language) %>%
  slice_max(order_by = n, n = 15) %>%
  pull(language)

movie_cuts %>%
  filter(language %in% top_languages) %>%
  bootstraps(times = 2000) %>%
  mutate(movie_data = map(splits, analysis)) %>%
  unnest(movie_data) %>%
  group_by(language, id) %>%
  summarize(avg_cuts = mean(n_cuts), .groups = "drop") %>%
  group_by(language) %>%
  summarize(
    mean = mean(avg_cuts),
    conf.low = quantile(avg_cuts, 0.025),
    conf.high = quantile(avg_cuts, 0.975)
  ) %>%
  mutate(language = fct_reorder(language, mean)) %>%
  ggplot(aes(x = mean, y = language)) +
  geom_vline(
    xintercept = mean(movie_cuts$n_cuts),
    linetype = "dashed",
    color = "gray50"
  ) +
  geom_errorbarh(
    aes(xmin = conf.low, xmax = conf.high),
    height = 0.2,
    color = "gray70"
  ) +
  geom_point(
    aes(color = mean > mean(movie_cuts$n_cuts)),
    size = 3
  ) +
  scale_color_manual(guide = "none", values = c("FALSE" = "#0072B2", "TRUE" = "#D55E00")) +
  labs(
    title = "Which Languages Receive More Censor Cuts?",
    subtitle = "95% confidence intervals for the average number of cuts per film.",
    x = "Average Cuts per Movie",
    y = "Language"
  ) +
  theme_minimal(base_family = "sans") +
  theme(
    panel.grid.major.y = element_blank()
  )

6 Languages certified by each office

Just curious about which offices modify which languages

valid_offices <- c("Mumbai", "Kolkata", "Hyderabad", "Guwahati", "Delhi",
                   "Cuttack", "Chennai", "Bangalore", "Thiruvananthpuram")

top_n_languages <- 15 

data %>%
  mutate(
    office = str_trim(office),
    office = str_to_title(office)
  ) %>%
    distinct(id, language, office) %>%
    filter(office %in% valid_offices, !is.na(language), language != "") %>%
    count(office, language, name = "count") %>%
    mutate(language = fct_lump_n(language, n = top_n_languages, w = count)) %>%
    filter(language != "Other") %>% 
    mutate(
    office = fct_reorder(office, count, .fun = sum),
    language = fct_reorder(language, count, .fun = sum, .desc = TRUE)
  ) %>%
  
  ggplot(aes(x = language, y = office, fill = count)) +
  geom_tile(color = "gray20", linewidth = 0.4) +
  
  geom_text(
    aes(label = count, color = count > 1200), 
    size = 2.75,
    fontface = "bold"
  ) +
  scale_fill_viridis_c(
    option = "magma",
    name = "Movie Count",
    direction = -1
  ) +
  scale_color_manual(guide = "none", values = c("FALSE" = "black", "TRUE" = "white")) +
  labs(
    title = paste("Frequency of Film Language by Certifying Office"),
    x = "",
    y = ""
  ) +
  theme_minimal(base_family = "sans") +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold", size = 16),
    plot.subtitle = element_text(hjust = 0.5, size = 12, color = "gray40"),
    axis.text.x = element_text(angle = 45, hjust = 1, vjust = 1),
    panel.grid = element_blank(),
    legend.position = "none"
  )