Scaling Labour 21 Conference Speeches

lab_fish.R

require(quanteda)
require(readtext)
require(tidyverse)
require(quanteda.textmodels)
require(quanteda.textplots)

path_data <- system.file("/Users/danbraby/labour2021/", package = "readtext")

# first save txt files from Github
dat <- readtext("/Users/danbraby/labour2021/txt/*",
                docvarsfrom = "filenames",
                docvarnames = c("name", "date"),
                dvsep = "_",
                encoding = "latin1")

# Read in flat-file from Github repo

# dat <- read.csv("https://raw.githubusercontent.com/dbrby/lab21_conference/main/lab21_speeches.csv")

dat_corp <- corpus(dat, docid_field = "name", text_field = "text")


dat_toks <- tokens(dat_corp, remove_punct = T,
                   remove_symbols = T, remove_numbers = T,
                   remove_url = T, remove_separators = T) %>%
  tokens_remove(stopwords("en")) %>%
  tokens_select(min_nchar = 2L)

dat_dfm <- dfm(dat_toks) %>% dfm_trim(min_docfreq = 2)


tmod_wf <- textmodel_wordfish(dat_dfm, dir = c(4, 5))

textplot_scale1d(tmod_wf)

textplot_scale1d(tmod_wf, margin = "features",
                 highlighted = c("fuel", "hgv",
                                 "aid", "rights",
                                 "women", "scotland",
                                 "keir", "tories",
                                 "boris"))