An R package to analyze the parliamentary records of the 19th legislative period of the Bundestag, the German parliament.
選択できるのは25トピックまでです。 トピックは、先頭が英数字で、英数字とダッシュ('-')を使用した35文字以内のものにしてください。

254 行
9.6KB

  1. #' Parse xml records
  2. #'
  3. #' Creates a list of tibbles containing relevant information from all records
  4. #' stored in the input directory.
  5. #'
  6. #' @param path path to records directory
  7. #' @param pattern search pattern to find records in directory
  8. #'
  9. #' @export
  10. read_all <- function(path="inst/records/", pattern="-data\\.xml") {
  11. # append file separator if needed
  12. path <- make_directory_path(path)
  13. cat("Reading all records from", path, "\n")
  14. # list all files in directory and filter by search pattern
  15. fs <- list.files(path)
  16. available_protocols <- fs[str_detect(fs, pattern)]
  17. if (length(available_protocols) == 0)
  18. stop(paste0("The given directory does not exist or does not contain files matching \"",
  19. pattern,
  20. "\"."))
  21. # parse records one by one and remove null entries
  22. res <- compact %$% pblapply(available_protocols, read_one, path=path)
  23. if (length(res) == 0) stop("No valid records found. Did you fetch successfully?")
  24. lapply(res, `[[`, "speaker") %>%
  25. bind_rows() %>%
  26. distinct() ->
  27. speaker
  28. lapply(res, `[[`, "speeches") %>%
  29. bind_rows() %>%
  30. distinct() %>%
  31. mutate(date = as.Date(date, format="%d.%m.%Y")) ->
  32. speeches
  33. lapply(res, `[[`, "talks") %>%
  34. bind_rows() %>%
  35. distinct() ->
  36. talks
  37. lapply(res, `[[`, "comments") %>%
  38. bind_rows() %>%
  39. distinct() ->
  40. commentsandapplause
  41. filter(commentsandapplause, type == "comment") %>%
  42. select(-type) ->
  43. comments
  44. filter(commentsandapplause, type == "applause") %>%
  45. select(-type, -commenter, -content) %>%
  46. mutate("CDU_CSU" = str_detect(fraction, "CDU/CSU"),
  47. "SPD" = str_detect(fraction, "SPD"),
  48. "FDP" = str_detect(fraction, "FDP"),
  49. "DIE_LINKE" = str_detect(fraction, "DIE LINKE"),
  50. "BUENDNIS_90_DIE_GRUENEN" = str_detect(fraction, "BÜNDNIS 90/DIE GRÜNEN"),
  51. "AfD" = str_detect(fraction, "AfD")) %>%
  52. select(-fraction) ->
  53. applause
  54. list(speaker = speaker, speeches = speeches, talks = talks, comments = comments, applause = applause)
  55. }
  56. # this reads all currently parseable data from one xml
  57. read_one <- function(name, path) {
  58. x <- tryCatch(read_xml(paste0(path, name)),
  59. error = function(c) NULL,
  60. warning = function(c) NULL)
  61. if (is.null(x)) return(NULL)
  62. # extract date of session
  63. date <- xml_attr(x, "sitzung-datum")
  64. cs <- xml_children(x)
  65. verlauf <- xml_find_first(x, "sitzungsverlauf")
  66. speakerl <- xml_find_first(x, "rednerliste")
  67. xml_children(speakerl) %>%
  68. parse_speakerlist() ->
  69. speaker
  70. xml_children(verlauf) %>%
  71. xml_find_all("rede") %>%
  72. parse_speechlist(date) ->
  73. res
  74. list(speaker = speaker, speeches = res$speeches, talks = res$talks, comments = res$comments)
  75. }
  76. xml_get <- function(node, name) {
  77. res <- xml_text %$% xml_find_all(node, name)
  78. if (length(res) == 0) NA_character_
  79. else res
  80. }
  81. # parse one speaker
  82. parse_speaker <- function(speaker_xml) {
  83. speaker_id <- xml_attr(speaker_xml, "id")
  84. nm <- xml_child(speaker_xml)
  85. prename <- xml_get(nm, "vorname")
  86. lastname <- xml_get(nm, "nachname")
  87. fraction <- xml_get(nm, "fraktion")
  88. title <- xml_get(nm, "titel")
  89. role <- xml_find_all(nm, "rolle")
  90. if (length(role) > 0) {
  91. role_long <- xml_get(role, "rolle_lang")
  92. role_short <- xml_get(role, "rolle_kurz")
  93. } else role_short <- role_long <- NA_character_
  94. c(id = speaker_id, prename = prename, lastname = lastname, fraction = fraction, title = title,
  95. role_short = role_short, role_long = role_long)
  96. }
  97. # parse one speech
  98. # returns: - a speech (with speech id and speaker id)
  99. # - all talks appearing in the speech (with corresponding content)
  100. parse_speech <- function(speech_xml, date) {
  101. speech_id <- xml_attr(speech_xml, "id")
  102. cs <- xml_children(speech_xml)
  103. cur_speaker <- NA_character_
  104. principal_speaker <- NA_character_
  105. cur_content <- ""
  106. speeches <- list()
  107. comments <- list()
  108. for (node in cs) {
  109. if (xml_name(node) == "p" || xml_name(node) == "name") {
  110. klasse <- xml_attr(node, "klasse")
  111. if ((!is.na(klasse) && klasse == "redner") || xml_name(node) == "name") {
  112. if (!is.na(cur_speaker)) {
  113. speech <- c(speech_id = speech_id,
  114. speaker = cur_speaker,
  115. content = cur_content)
  116. speeches <- c(speeches, list(speech))
  117. cur_content <- ""
  118. }
  119. if (is.na(principal_speaker) && xml_name(node) != "name") {
  120. principal_speaker <- xml_child(node) %>% xml_attr("id")
  121. }
  122. if (xml_name(node) == "name") {
  123. cur_speaker <- "BTP"
  124. } else {
  125. cur_speaker <- xml_child(node) %>% xml_attr("id")
  126. }
  127. } else {
  128. cur_content <- paste0(cur_content, xml_text(node), sep="\n")
  129. }
  130. } else if (xml_name(node) == "kommentar") {
  131. # comments are of the form
  132. # <kommentar>(blabla [Fraktion] – blabla liasdf – bla)</kommentar>
  133. xml_text(node) %>%
  134. str_sub(2, -2) %>%
  135. str_split("–") %>%
  136. `[[`(1) %>%
  137. lapply(parse_comment, speech_id = speech_id, on_speaker = cur_speaker) ->
  138. cs
  139. comments <- c(comments, cs)
  140. }
  141. }
  142. speech <- c(speech_id = speech_id,
  143. speaker = cur_speaker,
  144. content = cur_content)
  145. speeches <- c(speeches, list(speech))
  146. list(speech = c(id = speech_id, speaker = principal_speaker, date = date),
  147. parts = speeches,
  148. comments = comments)
  149. }
  150. fractionpattern <- "BÜNDNIS(SES)?\\W*90/DIE\\W*GRÜNEN|CDU/CSU|AfD|SPD|DIE LINKE|FDP|LINKEN"
  151. fractionnames <- c("BÜNDNIS 90/DIE GRÜNEN", "CDU/CSU", "AfD", "SPD", "DIE LINKE", "FDP")
  152. parse_comment <- function(comment, speech_id, on_speaker) {
  153. base <- c(speech_id = speech_id, on_speaker = on_speaker)
  154. # classify comment
  155. if(str_detect(comment, "Beifall")) {
  156. str_extract_all(comment, fractionpattern) %>%
  157. `[[`(1) %>%
  158. sapply(partial(flip(head), 1) %.% agrep, x=fractionnames, max=0.2, value=T) %>%
  159. str_c(collapse=",") ->
  160. by
  161. c(base, type = "applause", fraction = by, commenter = NA_character_, content = comment)
  162. } else {
  163. ps <- str_match(comment, "(.*) \\[(.*?)\\]: (.*)")[1,]
  164. c(base, type = "comment", fraction = ps[3], commenter = ps[2], content = ps[4])
  165. }
  166. }
  167. # creates a tibble of speeches and a tibble of talks from a list of xml nodes representing speeches
  168. parse_speechlist <- function(speechlist_xml, date) {
  169. d <- sapply(speechlist_xml, parse_speech, date = date)
  170. speeches <- simplify2array(d["speech", ])
  171. parts <- simplify2array %$% unlist(d["parts", ], recursive=FALSE)
  172. comments <- simplify2array %$% unlist(d["comments", ], recursive=FALSE)
  173. list(speeches = tibble(id = speeches["id",], speaker = speeches["speaker",],
  174. date = speeches["date",]),
  175. talks = tibble(speech_id = parts["speech_id", ],
  176. speaker = parts["speaker", ],
  177. content = parts["content", ]),
  178. comments = tibble(speech_id = comments["speech_id",],
  179. on_speaker = comments["on_speaker",],
  180. type = comments["type",],
  181. fraction = comments["fraction",],
  182. commenter = comments["commenter",],
  183. content = comments["content", ]))
  184. }
  185. # create a tibble of speaker from a list of xml nodes representing speaker
  186. parse_speakerlist <- function(speakerliste_xml) {
  187. d <- sapply(speakerliste_xml, parse_speaker)
  188. tibble(id = d["id",],
  189. prename = d["prename",],
  190. lastname = d["lastname",],
  191. fraction = d["fraction",],
  192. title = d["title",],
  193. role_short = d["role_short",],
  194. role_long = d["role_long",])
  195. }
  196. #' Write the parsed and repaired results into separate csv files
  197. #'
  198. #' @param tables list of tables to convert into a csv files.
  199. #' @param path where to put the csv files.
  200. #' @param create set TRUE if the path does not exist yet and you want to create it
  201. #'
  202. #' @export
  203. write_to_csv <- function(tables, path="inst/csv/", create=F) {
  204. check_directory(path, create)
  205. write.table(tables$speaker, str_c(path, "speaker.csv"))
  206. write.table(tables$speeches, str_c(path, "speeches.csv"))
  207. write.table(tables$talks, str_c(path, "talks.csv"))
  208. write.table(tables$comments, str_c(path, "comments.csv"))
  209. write.table(tables$applause, str_c(path, "applause.csv"))
  210. }
  211. #' create a tibble from the csv file
  212. #'
  213. #' @param path directory to read files from
  214. #'
  215. #' reading the tables from a csv is way faster than reading and repairing the data every single time
  216. #'
  217. #' @export
  218. read_from_csv <- function(path="inst/csv/") {
  219. list(speaker = read.table(str_c(path, "speaker.csv")) %>%
  220. tibble() %>%
  221. mutate(id = as.character(id)),
  222. speeches = read.table(str_c(path, "speeches.csv")) %>%
  223. tibble() %>%
  224. mutate(speaker = as.character(speaker),
  225. date = as.Date(date)),
  226. talks = tibble %$% read.table(str_c(path, "talks.csv")),
  227. comments = tibble %$% read.table(str_c(path, "comments.csv")),
  228. applause = tibble %$% read.table(str_c(path, "applause.csv")))
  229. }