Mining the last French presidential debate

After reading this post (thanks to him), I think it could be interesting to replicate this with some specific up of french language and to see and we can perform rapid view of the debate between Sarkozy and Hollande of the last 2nd round of presidential election.

Key words : TextMining, Elections, France, Debate, 2nd Round

We use the packages qdap from (Tyler Rinker) and tm to perform textmining analysis and the classical package like ggplot or RColorBrewer make our  graphics look pretty.
For Hollande

	# 18/05/2013
	# Key words : TextMining, Elections, France, Debate, 2nd Round
	# We use the packages qdap from (donner le lien) and
	# tm to perform textmining analysis and the classical
	# package like ggplot or RColorBrewer to get the graphics pretty.
	suppressPackageStartupMessages(require(twitteR))
	suppressPackageStartupMessages(require(XML))
	suppressPackageStartupMessages(require(tm))
	suppressPackageStartupMessages(require(rgdal))
	suppressPackageStartupMessages(require(ggplot2))
	suppressPackageStartupMessages(require(qdap))
	suppressPackageStartupMessages(require(rJava))
	suppressPackageStartupMessages(library(wordcloud))
	library(Rstem)
	setwd("D:/PERSO/R_Working/Tutoriels/TextMining")
	# Hollande
	debate <- read.transcript("./Data/debat2tours.docx", col.names=c("person", "dialogue"))
	htruncdf(debate,5,50)
	# We keep just Holland's word
	Hollande = subset(debate,person=="HOLLANDE")
	# We define the stop words
	sw=c("a","ou",tm::stopwords("fr"),"c'est", "n'est","s'y","qu'on","s'il","ah",
	letters,"ca","n'y","d'un","monsieur")
	generateCorpus= function(df,my.stopwords=c()){
	text2.corpus= Corpus(VectorSource(df),readerControl=list(language="fr"))
	text2.corpus = tm_map(text2.corpus, removePunctuation)
	text2.corpus = tm_map(text2.corpus, tolower)
	text2.corpus= tm_map(text2.corpus, removeNumbers)
	text2.corpus = tm_map(text2.corpus, removeWords, stopwords("fr"))
	text2.corpus = tm_map(text2.corpus, removeWords, my.stopwords)
	#text2.corpus <- tm_map(text2.corpus, stemDocument, language = "french")
	}
	HollandeCorpus<-generateCorpus(Hollande,sw)
	# We build a Term Document Matrix
	H.tdm <- TermDocumentMatrix(HollandeCorpus)
	H.m <- as.matrix(H.tdm)
	H.v <- sort(rowSums(H.m),decreasing=TRUE)
	H.d <- data.frame(word = names(H.v),freq=H.v)
	H.d = subset(H.d,freq<=90)
	H.d = subset(H.d,freq>=3)
	H.d$stem <- wordStem(row.names(H.d), language = "french")
	# and put words to column, otherwise they would be lost when aggregating
	H.d$word <- row.names(H.d)
	agg_freq <- stats::aggregate(freq ~ stem, data = H.d, sum)
	agg_word <- stats::aggregate(word ~ stem, data = H.d, function(x) x[1])
	forW <- cbind(freq = agg_freq[, 2], agg_word)
	# sort by frequency
	forW <- forW[order(forW$freq, decreasing = T), ]
	# Wordcloud
	col<- brewer.pal(8,"Dark2")
	png("wordcloud_Hollande.png", width=1280,height=800)
	wordcloud(forW$word,forW$freq, scale=c(8,.2),min.freq=5,
	max.words=Inf, random.order=FALSE, rot.per=.20, colors=col)
	dev.off()

view raw MiningFrenchPresidentialdebate.r hosted with ❤ by GitHub

Top words From hollande

For Sarkozy

Top words from Sarkozy

As we can see, I've a problem to manage french accent. If somebody have any idea... We can also perform a quick Gantt plot basing on Qdap package and get some information about who lead the debate

	debate2 = subset(debate,person=="SARKOZY"|person == "HOLLANDE")
	debate2$person<- factor(debate2$person,levels=qcv(terms="SARKOZY HOLLANDE"))
	png("Gant.png", width=700,height=500)
	with(debate2, gantt_plot(dialogue, person, xlab = "duration(words)", scale = "free"))
	dev.off()

view raw Miningdebate2.r hosted with ❤ by GitHub

No surprise about the winner.