UN_text_analysis_ORF.Rmd

---
title: "UNGA-UNSC Article"
author: "Jayati Sharma"
date: "2023-01-18"
output: html_document
---

```{r setup, include=FALSE}
library(readtext) #handling text files
library(dplyr) #data manipulation
library(tidyverse) #for data wrangling and data manipulation
library(ggplot2) #data visualizations
library(viridis) #for colours
library(RColorBrewer) #for colors
library(readxl) #importing excel files
library(stringr) #string manipulation
library(ggthemes) #for visualization themes
library(classInt) #choosing univariate class intervals
library(quanteda) #for textual analysis
library(quanteda.textplots) #plotting functions for visualizing textual data
library(quanteda.textstats) #for textual statistics
library(tokenizers) #convert natural language text into tokens
library(furrr)

#setting directory
DATA_DIR <- ""
```

```{r import files, include = FALSE}

#Reading in text files
ungd_files <- readtext(paste0(DATA_DIR, "//*"),
                                 docvarsfrom = "filenames", 
                                 dvsep= "_", 
                                 docvarnames = c("Country", "Session", "Year"))

#changing row.names to have only country_year, rather than folder pathway from `readtext`.

 ungd_files$doc_id <- str_replace(ungd_files$doc_id , ".txt", "") %>%
   str_replace(. , "_\\d{2}", "")
``` 

```{r Creating corpus, include=FALSE}

#Creating corpus of speeches
ungd_corpus <- corpus(ungd_files, text_field = "text")
```

```{r Converting into tokens, echo=FALSE}

#converting speeches into tokens
  token_all_speeches <- tokens(ungd_corpus,what = "word",remove_punct = TRUE, remove_symbols = TRUE,
  remove_numbers = TRUE, remove_url = TRUE, remove_separators = TRUE,
  split_hyphens = TRUE, include_docvars = FALSE, padding = FALSE)

#removing stopwords and converting to lowercase
  token_all_speeches <- tokens_remove(token_all_speeches, c(stopwords("english")))
  token_all_speeches <- tokens_tolower(token_all_speeches)
```

```{r all words , echo = FALSE}

#making a document-feature matrix from the tokens
dfm_token_all_speeches <- dfm(
token_all_speeches,
tolower = TRUE,
remove_padding = FALSE,
verbose = quanteda_options("verbose"))

#creating a dataframe with frequency of each word spoken in each year
freq_words_all_speeches <- textstat_frequency(dfm_token_all_speeches, groups = ungd_files$Year)
```

```{r terrorism, echo=FALSE}

#creating a subset of the dataframe for words related to terrorism
freq_terrorism <- subset(freq_words_all_speeches, freq_words_all_speeches$feature %in% c("terrorism", "terrorist", "terrorists"))

#calculating frequency
terrorism_words <- freq_terrorism %>%
  group_by(group) %>% summarise(Times = sum(frequency))

#plotting it
ggplot(terrorism_words, aes(x = group, y = Times)) +
    xlab("Year") + 
    ylab("Frequency") +
  geom_line(group = 1, color = '#009cde', lwd = 1) +
  ggtitle("Mentions of Terrorism Peaked at the Time of 9/11",subtitle = "Number of Times Leaders Mentioned Terrorism in UNGD Speeches Between 1970-2020") +
  geom_point(shape = 1, size = 1.5)+
  labs(caption = "Data Source: United Nations General Debate Corpus by Jankin Mikhaylov, Slava and Baturo, Alexander and Dasandi, Niheer ; Analysis by CPC Analytics")+
  scale_x_discrete(breaks = seq(1970, 2020, by= 5))+
  theme_fivethirtyeight()

ggsave(filename = "terrorism_words.jpg", width = 10, height = 4)

#looking at the context in which terrorism was spoken of
context_terrorism <- kwic(token_all_speeches, c("terrorism", "terrorist", "terrorists"), window = 4)
context_terrorism <- cbind(context_terrorism, str_split_fixed(context_terrorism$docname, "_", 2))

#looking at the countries which spoke about terrorism
country_speaking_terrorism <- context_terrorism %>%
  group_by(`1`) %>%
  summarise(country_terrorism_freq = n())
```

```{r refugees}

#creating a subset of the dataframe for words related to refugees

freq_refugees <- subset(freq_words_all_speeches, freq_words_all_speeches$feature %in% c("refugee", "refugees"))

#calcualting frequency
refugees_words <- freq_refugees %>%
  group_by(group) %>% summarise(Times = sum(frequency))

#plotting it
ggplot(refugees_words, aes(x = group, y = Times)) +
    xlab("Year") + 
    ylab("Frequency") +
  geom_line(group = 1, color = '#009cde', lwd = 1) +
  ggtitle("Mentions of Refugees Peaked during 2016 Refugee Crisis",subtitle = "Number of Times Leaders Mentioned Refugees in UNGD Speeches Between 1970-2020") +
  geom_point(shape = 1, size = 1.5)+
  labs(caption = "Data Source: United Nations General Debate Corpus by Jankin Mikhaylov, Slava and Baturo, Alexander and Dasandi, Niheer ; Analysis by CPC Analytics")+
  scale_x_discrete(breaks = seq(1970, 2020, by= 5))+
  theme_fivethirtyeight()
ggsave(filename = "refugees_words.jpg", width = 10, height = 4)

#looking at the context in which it was spoken
context_refugees <- kwic(token_all_speeches, c("refugee", "refugees"), window = 4)
context_refugees <- cbind(context_refugees, str_split_fixed(context_refugees$docname, "_", 2))

#frequency of countries that mentioned it
country_speaking_refugees <- context_refugees %>%
  group_by(`1`) %>%
  summarise(country_refugees_freq = n())
```

```{r climate change, echo=FALSE}

#joining words where climate is succeeded by change
climate_pattern <- c("climate change")
climate_token <- tokens_compound(token_all_speeches,pattern = phrase(climate_pattern), join = TRUE, concatenator = "_",)

# a document-feature matrix of the new speeches with climate_change joined 
climate_dfm_token_all_speeches <- dfm(
climate_token,
tolower = TRUE,
remove_padding = FALSE,
verbose = quanteda_options("verbose"))

#frequency of all words
climate_freq_words_all_speeches <- textstat_frequency(climate_dfm_token_all_speeches, groups = ungd_files$Year)

#creating a subset of the dataframe for words related to climate change
freq_climate_change <- subset(climate_freq_words_all_speeches, climate_freq_words_all_speeches$feature %in% c("climate_change"))

#calculating the frequency
climate_change_words <- freq_climate_change %>%
  group_by(group) %>% summarise(Times = sum(frequency))

#plotting it
ggplot(climate_change_words, aes(x = group, y = Times)) +
    xlab("Year") + 
    ylab("Frequency") +
  geom_line(group = 1, color = '#009cde', lwd = 1) +
  ggtitle("Recent Rise in the Attention to Climate Change",subtitle = "Number of Times Leaders Mentioned Climate Change in UNGD Speeches between 1970-2020") +
  geom_point(shape = 1, size = 1.5)+
  labs(caption = "Data Source: United Nations General Debate Corpus by Jankin Mikhaylov, Slava and Baturo, Alexander and Dasandi, Niheer ; Analysis by CPC Analytics")+
  scale_x_discrete(breaks = seq(1970, 2020, by= 5))+
  theme_fivethirtyeight()
ggsave(filename = "climatechange_words.jpg", width = 10, height = 5)

#looking at the context in which the word was spoken
context_climate_change <- kwic(climate_token, c("climate_change"), window = 6)
context_climate_change <- cbind(context_climate_change, str_split_fixed(context_climate_change$docname, "_", 2))

#frequency of the countries mentioning it
country_speaking_climate <- context_climate_change %>%
  group_by(`1`) %>%
  summarise(country_climate_freq = n())
```

```{r disarmament, echo=FALSE}

#creating a subset of the dataframe for words related to disarmament

freq_disarmament <- subset(freq_words_all_speeches, freq_words_all_speeches$feature %in% c("disarmament"))

#calculating the frequency
disarmament_words <- freq_disarmament %>%
  group_by(group) %>% summarise(Times = sum(frequency))

#plotting it
ggplot(disarmament_words, aes(x = group, y = Times)) +
    xlab("Year") + 
    ylab("Frequency") +
  geom_line(group = 1, color = '#009cde', lwd = 1) +
  ggtitle("The Disarmament Agenda has Lost Momentum Over the Years",subtitle = "Number of Times Leaders Mentioned Disarmament in UNGD Speeches Between 1970-2020") +
  geom_point(shape = 1, size = 1.5)+
  labs(caption = "Data Source: United Nations General Debate Corpus by Jankin Mikhaylov, Slava and Baturo, Alexander and Dasandi, Niheer ; Analysis by CPC Analytics")+
  scale_x_discrete(breaks = seq(1970, 2020, by= 5))+
  theme_fivethirtyeight()
ggsave(filename = "disarmamentwords_words.jpg", width = 10, height = 4)

#context in which the word was spoken
context_disarmament <- kwic(token_all_speeches, c("disarmament"), window = 4)
context_disarmament <- cbind(context_disarmament, str_split_fixed(context_disarmament$docname, "_", 2))

#frequency of countries mentioning it
country_speaking_disarmament <- context_disarmament %>%
  group_by(`1`) %>%
  summarise(country_disarmament_freq = n())
```