## word frequency matrix (wfm) example:
with(DATA, wfm(state, list(sex, adult)))[1:15, ]
with(DATA, wfm(state, person))[1:15, ]
Filter(with(DATA, wfm(state, list(sex, adult))), 5)
with(DATA, wfm(state, list(sex, adult)))
## Filter particular words based on max/min values in wfm
v <- with(DATA, wfm(state, list(sex, adult)))
Filter(v, 5)
Filter(v, 5, count.apostrophe = FALSE)
Filter(v, 5, 7)
Filter(v, 4, 4)
Filter(v, 3, 4)
Filter(v, 3, 4, stopwords = Top25Words)
## insert double tilde ("~~") to keep phrases(i.e., first last name)
alts <- c(" fun", "I ")
state2 <- space_fill(DATA$state, alts, rm.extra = FALSE)
with(DATA, wfm(state2, list(sex, adult)))[1:18, ]
## word frequency dataframe (wfdf) example:
with(DATA, wfdf(state, list(sex, adult)))[1:15, ]
with(DATA, wfdf(state, person))[1:15, ]
## wfm_expanded example:
z <- wfm(DATA$state, DATA$person)
wfm_expanded(z)[30:45, ] #two "you"s
## wf_combine examples:
#===================
## raw no margins (will work)
x <- wfm(DATA$state, DATA$person)
## raw with margin (will work)
y <- wfdf(DATA$state, DATA$person, margins = TRUE)
## Proportion matrix
z2 <- wfm(DATA$state, DATA$person, output="proportion")
WL1 <- c(y[, 1])
WL2 <- list(c("read", "the", "a"), c("you", "your", "you're"))
WL3 <- list(bob = c("read", "the", "a"), yous = c("you", "your", "you're"))
WL4 <- list(bob = c("read", "the", "a"), yous = c("a", "you", "your", "your're"))
WL5 <- list(yous = c("you", "your", "your're"))
WL6 <- list(c("you", "your", "your're")) #no name so will be called words 1
WL7 <- c("you", "your", "your're")
wfm_combine(z, WL2) #Won't work not a raw frequency matrix
wfm_combine(x, WL2) #Works (raw and no margins)
wfm_combine(y, WL2) #Works (raw with margins)
wfm_combine(y, c("you", "your", "your're"))
wfm_combine(y, WL1)
wfm_combine(y, WL3)
## wfm_combine(y, WL4) #Error
wfm_combine(y, WL5)
wfm_combine(y, WL6)
wfm_combine(y, WL7)
worlis <- c("you", "it", "it's", "no", "not", "we")
y <- wfdf(DATA$state, list(DATA$sex, DATA$adult), margins = TRUE)
z <- wfm_combine(y, worlis)
chisq.test(z)
chisq.test(wfm(y))
## Dendrogram
presdeb <- with(pres_debates2012, wfm(dialogue, list(person, time)))
library(sjPlot)
sjc.dend(t(presdeb), 2:4)
## Words correlated within turns of talk
## EXAMPLE 1
library(reports)
x <- factor(with(rajSPLIT, paste(act, pad(TOT(tot)), sep = "|")))
dat <- wfm(rajSPLIT$dialogue, x)
cor(t(dat)[, c("romeo", "juliet")])
cor(t(dat)[, c("romeo", "banished")])
cor(t(dat)[, c("romeo", "juliet", "hate", "love")])
qheat(cor(t(dat)[, c("romeo", "juliet", "hate", "love")]),
diag.na = TRUE, values = TRUE, digits = 3, by.column = NULL)
dat2 <- wfm(DATA$state, id(DATA))
qheat(cor(t(dat2)), low = "yellow", high = "red",
grid = "grey90", diag.na = TRUE, by.column = NULL)
## EXAMPLE 2
x2 <- factor(with(pres_debates2012, paste(time, pad(TOT(tot)), sep = "|")))
dat2 <- wfm(pres_debates2012$dialogue, x2)
wrds <- word_list(pres_debates2012$dialogue,
stopwords = c("it's", "that's", Top200Words))
wrds2 <- tolower(sort(wrds$rfswl[[1]][, 1]))
qheat(word_cor(t(dat2), word = wrds2, r = NULL),
diag.na = TRUE, values = TRUE, digits = 3, by.column = NULL,
high="red", low="yellow", grid=NULL)
## EXAMPLE 3
library(gridExtra); library(ggplot2); library(grid)
dat3 <- lapply(qcv(OBAMA, ROMNEY), function(x) {
with(pres_debates2012, wfm(dialogue[person == x], x2[person == x]))
})
# Presidential debates by person
dat5 <- pres_debates2012
dat5 <- dat5[dat5$person %in% qcv(ROMNEY, OBAMA), ]
disp <- with(dat5, dispersion_plot(dialogue, wrds2, grouping.var = person,
total.color = NULL, rm.vars=time))
cors <- lapply(dat3, function(m) {
word_cor(t(m), word = wrds2, r = NULL)
})
plots <- lapply(cors, function(x) {
qheat(x, diag.na = TRUE, values = TRUE, digits = 3, plot = FALSE,
by.column = NULL, high="red", low="yellow", grid=NULL)
})
plots <- lapply(1:2, function(i) {
plots[[i]] + ggtitle(qcv(OBAMA, ROMNEY)[i]) +
theme(axis.title.x = element_blank(),
plot.margin = unit(rep(0, 4), "lines"))
})
grid.arrange(disp, arrangeGrob(plots[[1]], plots[[2]], ncol=1), ncol=2)
## With `word_cor`
worlis <- list(
pronouns = c("you", "it", "it's", "we", "i'm", "i"),
negative = qcv(no, dumb, distrust, not, stinks),
literacy = qcv(computer, talking, telling)
)
y <- wfdf(DATA$state, id(DATA, prefix = TRUE))
z <- wfm_combine(y, worlis)
word_cor(t(z), word = names(worlis), r = NULL)
## Plotting method
plot(y, TRUE)
plot(z)
## Correspondence Analysis
library(ca)
dat <- pres_debates2012
dat <- dat[dat$person %in% qcv(ROMNEY, OBAMA), ]
speech <- stemmer(dat$dialogue)
mytable1 <- with(dat, wfm(speech, list(person, time), stopwords = Top25Words))
fit <- ca(mytable1)
summary(fit)
plot(fit)
plot3d.ca(fit, labels=1)
mytable2 <- with(dat, wfm(speech, list(person, time), stopwords = Top200Words))
fit2 <- ca(mytable2)
summary(fit2)
plot(fit2)
plot3d.ca(fit2, labels=1)
## Weight a wfm
WFM <- with(DATA, wfm(state, list(sex, adult)))
plot(weight(WFM, "scaled"), TRUE)
weight(WFM, "prop")
weight(WFM, "max")
weight(WFM, "scaled")Run the code above in your browser using DataLab