Analysis of Dason's analysis of Lil Wayne

gistfile1.r

pstat <- require(pacman)
bostat <- require(BreakoutDetection)
if (!pstat) devtools::install_github("trinker/pacman")
if (!bostat) devtools::install_github("twitter/BreakoutDetection")
p_load(BreakoutDetection, dplyr, XML, qdap)

doc <- htmlParse("http://dasonk.com/2014/07/18/Lil-Jon-Analysis/")
text <- read.transcript(text = head(sapply(getNodeSet(doc, "//p"), xmlValue), -1))
sents <- sent_detect(text[[1]])

dpol <- polarity(sents)
plot(dpol)
cumpolarity <- cumulative(dpol)

res2 <- breakout(cumpolarity[[1]], min.size=4, method='multi', beta=.3, degree=0, plot=TRUE)

res2$plot  ## ggplot
dev.new()
cumpolarity

cumpolarity_2 <- data.frame(
    cum_mean = cumpolarity[[1]], 
    Time = seq_along(cumpolarity[[1]])
)
polcount <- counts(dpol)[["polarity"]]

ends2 <- c(res2[["loc"]], length(cumpolarity[[1]]))
starts2 <- c(1, head(ends2, -1)) 
rects2 <- data.frame(xstart = starts2, xend = ends2,
    period = c("Unstable", LETTERS[1:c(-1 + length(ends2))]))

var <- rep(NA, length(cumpolarity[[1]]))

for (i in 1:nrow(rects2)) {
    var[rects2[i, 1]:rects2[i, 2]] <- as.character(rects2[i,3])
}

rects2[["cumave"]] <- sapply(rects2[["xend"]], function(i) mean(cumpolarity[[1]][1:i]))

rects2[["ave"]] <- data.frame(group = var, 
    cummean = cumpolarity[[1]]) %>%
    group_by(group) %>%
    summarize(ave=mean(cummean)) %>%
    with(., lookup(rects2[["period"]], group, ave))

(dplot <- plot(cumpolarity, plot=FALSE)  +
    geom_rect(data = rects2, aes(xmin = xstart, xmax = xend,
        ymin = -Inf, ymax = Inf, fill = period), alpha = 0.17) +
    geom_text(data=rects2, aes(x=(xstart + xend)/2, y=-.04,
        label=period), size=3) +
    guides(fill=FALSE) +
    geom_vline(xintercept=c(res2[["loc"]]), alpha=.2, color="red") +
    theme(
        panel.grid.minor.x=element_blank(), 
        panel.grid.major.x=element_blank()
    ) +
    #xlim(0, length(cumpolarity[[1]])+5)+
    geom_segment(data = rects2, linetype = "longdash", 
        color="darkgreen", alpha=.4, size=1.25, 
        aes(x = xstart, y = ave, xend = xend, yend = ave)) +
    scale_x_continuous(expand = c(0,0), limits = c(0, length(cumpolarity[[1]])+1)))

dplot

out <- split(sents, var)[unique(var)]
dbot <- setNames(out, paste(names(out), round(rects2[["ave"]], 4), sep=": average polarity = "))
dbot