Exercises and Answers

  1. Change text_data_2 into a vector containing just word “article” and respective number in each element (using function str_extract()), then randomly change order of elements. Finally, order output vector using function str_order(). text_data_2 is avaliable in csv file exercise1.

Proposed answer:

library(readr)
library(stringr)
exercise1 <- read.csv("exercise1")$x
extract <- str_extract(string = exercise1, pattern = "Article \\d")
random_order <- extract[sample(1:length(extract), length(extract))]
indexes <- str_order(random_order)
random_order[indexes]
## [1] "Article 1" "Article 2" "Article 3" "Article 4" "Article 5"
  1. Change text_data_2 into a vector containing just word “article” and respective number in each element (using function str_remove_all()), then randomly change order of elements. Finally, order output vector using function str_sort(). text_data_2 is avaliable in csv file exercise1.

Proposed answer:

library(readr)
library(stringr)
exercise1 <- read.csv("exercise1")$x
remove1 <- str_remove_all(string = exercise1, pattern = "\n")
remove2 <- str_remove_all(string = remove1, pattern = "(?<=(Article \\d)).{1,}")
random_order <- remove2[sample(1:length(remove2), length(remove2))]
str_sort(random_order)
## [1] "Article 1" "Article 2" "Article 3" "Article 4" "Article 5"
  1. Download table of current members of the United Nations from these website and remove last column. From the first column remove notes (i.e. strings “[note 1]”, “[note 2]” and so on).

Proposed answer:

library(rvest)
library(stringr)
library(dplyr)
url <- "https://en.wikipedia.org/wiki/Member_states_of_the_United_Nations" 
url %>%
  read_html() %>%
  html_node(xpath='//*[@id="mw-content-text"]/div[1]/table[2]') %>%
  html_table(fill=TRUE) -> table
table <- table[,-3]
table$`Member state` <- str_remove(string = table$`Member state`, pattern = "\\[note \\d\\]")
table
## # A tibble: 193 × 2
##    `Member state`      `Date of admission`
##    <chr>               <chr>              
##  1 Afghanistan         19 November 1946   
##  2 Albania             14 December 1955   
##  3 Algeria             8 October 1962     
##  4 Andorra             28 July 1993       
##  5 Angola              1 December 1976    
##  6 Antigua and Barbuda 11 November 1981   
##  7 Argentina           24 October 1945    
##  8 Armenia             2 March 1992       
##  9 Australia           1 November 1945    
## 10 Austria             14 December 1955   
## # … with 183 more rows
  1. Using eurlex package download document with CELEX number 32021R2282, examine term frequency and plot the results as wordcloud. Plot 72 the most frequent words.

Proposed answer:

library(eurlex)
library(textstem)
library(tidytext)
library(wordcloud2)
library(dplyr)

elx_make_query(resource_type = "regulation",
               directory = "16") %>% 
  elx_run_query() -> dataset

dataset <- dataset[match("32021R2282", dataset$celex),]

legal.act <- elx_fetch_data(dataset$work,
                            type = "text",
                            language_1 = "en")

legal.act <- data.frame(text = legal.act)

legal.act %>%
  unnest_tokens(word, text, token = "ngrams", n = 1) %>%
  anti_join(stop_words) %>%
  mutate(word = lemmatize_words(word)) -> legal.act

legal.act$word <- ifelse(str_remove_all(legal.act$word, "\\d") == "",
                         NA,
                         str_remove_all(legal.act$word,"\\d"))

legal.act <- na.omit(legal.act)

legal.act %>%
  group_by(word) %>%
  summarise(count = n()) %>%
  arrange(desc(count)) %>%
  slice(1:72) -> legal.act.frequency

wordcloud2(legal.act.frequency)
  1. Using eurlex package download legal acts specified in csv file exercise5. Then examine what these documents are about and plot the results, but plot just six the most important words. Hint: directory is 1607.

Proposed answer:

library(readr)
library(eurlex)
library(tidytext)
library(stringr)
library(ggplot2)
library(textstem)
library(dplyr)

exercise5 <- read.csv("exercise5")$x

elx_make_query(resource_type = "any",
               directory = "1607") %>% 
  elx_run_query() -> dataset

dataset <- dataset[match(exercise5, dataset$celex),]

legal.acts <- data.frame(CELEX=dataset$celex,
                         text=unlist(lapply(dataset$work,elx_fetch_data,
                                            language_1 = "en",
                                            type = "text")))

legal.acts$text <- str_remove_all(legal.acts$text,"\\d")

legal.acts %>%
  unnest_tokens(word, text) %>%
  anti_join(stop_words) %>%
  mutate(word = lemmatize_words(word)) %>% 
  group_by(CELEX, word)%>%
  summarise(n = n())%>%
  arrange(desc(n)) -> act.word

act.word %>%
  bind_tf_idf(word, CELEX, n) -> act.word.2

act.word.2 %>% 
  arrange(desc(tf_idf)) %>%
  mutate(word = factor(word, levels = rev(unique(word)))) %>% 
  group_by(CELEX) %>% 
  slice(1:6) %>% 
  ungroup() %>%
  ggplot(aes(word, tf_idf, fill = CELEX)) +
  geom_col(show.legend = FALSE) +
  labs(x = NULL, y = "tf-idf") +
  facet_wrap(~CELEX, ncol = 2, scales = "free") +
  coord_flip()+
  theme_bw()

  1. Group documents from csv file exercise6. There should be six groups: “Regulations”, “Decisions”, “Recommendations”, “Opinions”, “Resolutions”, “Other documents”. Create data frame where the first column is CELEX number and the second column is group name.

Proposed answer:

Hello, Buddy! If you thought that you should have use LDA, then sorry, you were wrong. I’m little mean. The answer is nothing more than pattern recognition. ;)

Import data to R environment.

dataset <- read.csv("exercise6")$x
dataset <- str_remove(string = dataset, pattern = "CELEX number: ")
uniques <- !duplicated(str_sub(dataset, start = 6L, end = 6L))
coding <- str_sub(dataset, start = 6L, end = 6L)
unique_codes <- coding[uniques]

Here you can find information about coding in CELEX numbers

library(rvest)
library(stringr)
library(dplyr)
url <- "https://eur-lex.europa.eu/content/tools/TableOfSectors/types_of_documents_in_eurlex.html"
url %>%
  read_html() %>%
  html_node(xpath='//*[@id="content"]/div/div/div/div[3]/table') %>%
  html_table(fill=TRUE) -> table
codes <- table[table$X2 %in% unique_codes,]
codes <- codes[,-1]
codes$X3 <- str_remove(string = codes$X3, pattern = " \\(with or without addressee\\)")
codes$X3 <- str_remove(string = codes$X3, pattern = " published in OJ C")
names(codes) <- c("code", "object")

codes
## # A tibble: 6 × 2
##   code  object         
##   <chr> <chr>          
## 1 R     Regulations    
## 2 D     Decisions      
## 3 H     Recommendations
## 4 A     Opinions       
## 5 G     Resolutions    
## 6 Y     Other documents

The rest is quite easy.

decoding <- function(x)
{
  return(codes$object[match(x,codes$code)])
}

answer <- data.frame(CELEX = dataset, code = coding)

answer %>%
  mutate(object = decoding(code)) -> answer

answer <- answer[,-2]

answer
##             CELEX          object
## 1      32021R0101     Regulations
## 2      32021R0100     Regulations
## 3  32017A0608(01)        Opinions
## 4  32015A1205(01)        Opinions
## 5  32015A0228(01)        Opinions
## 6  32014A1122(01)        Opinions
## 7  32010A1222(01)        Opinions
## 8      32010H0635 Recommendations
## 9  32010A0928(01)        Opinions
## 10 32010A0804(01)        Opinions
## 11 32010A0415(01)        Opinions
## 12 32010A0414(01)        Opinions
## 13 32010A0320(01)        Opinions
## 14 32010A0318(01)        Opinions
## 15 32009A0707(01)        Opinions
## 16     32009H0120 Recommendations
## 17 32008A0509(01)        Opinions
## 18 22008A0208(01)        Opinions
## 19 32007A1221(01)        Opinions
## 20     32007D0513       Decisions
## 21     32006D0908       Decisions
## 22     32006D0626       Decisions
## 23     32006H0040 Recommendations
## 24     32005R0302     Regulations
## 25 22002A1127(01)        Opinions
## 26 32002G0522(01)     Resolutions
## 27 31999Y0209(01) Other documents
## 28     31997D0873       Decisions
## 29     31996D0671       Decisions
## 30     31994D0955       Decisions
## 31     31994H0956 Recommendations
## 32 21994A0803(01)        Opinions
## 33     31993R1493     Regulations
## 34 31992Y0708(02) Other documents
## 35     31992D0194       Decisions
## 36     31990D0413       Decisions
## 37     31989A0476        Opinions
## 38     31989A0354        Opinions
## 39     31989A0082        Opinions
## 40 31975Y0814(01) Other documents