Aplikace v R pro Corpus Viewer

server.R

library(shiny)
library(RCurl)
library(RJSONIO)
library(textcat)
library(tm)
library(tau)
library(corrplot)

# Define server logic 
shinyServer(function(input, output) {
  
  
  # load corpus from jobs publishers (sample set from Havel speechs)
  obj.temp <- read.csv("prace-full.csv")
  corpus.prace <- as.vector(obj.temp$POZNAMKA[1:1000])
  corpus.prace <- Corpus(VectorSource(corpus.prace))
  corpus.prace <- tm_map(corpus.prace, tolower)
  corpus.prace <- tm_map(corpus.prace, removePunctuation)
  corpus.prace <- tm_map(corpus.prace, removeNumbers)
  dtm.prace <- DocumentTermMatrix(corpus.prace)
  
  # create reactive function for switching between datasets  
  datasetInput <- reactive(function() {
     corpus.prace
  })


  # create reactive function for switching between document term matrix  
  termetrixinput <- reactive(function() {
      dtm.prace
  })

  # create reactive function for sparsing  
  dtmInput <- reactive(function() {
    dtm <- termetrixinput()
    dtm <- removeSparseTerms(dtm, input$sparse)  
  })

  
  # quick summary for whole corpus
  output$numDocuments <- reactivePrint(function() {
    dataset <- datasetInput() 
    summary(dataset)
  })

  # calculate N-grams for whole corpus   
  output$viewNgram <- reactivePrint(function() {
    dataset <- datasetInput() 
    kindofngrams <- textcnt(dataset, method = "string",n=input$numngrams)
    data.frame(count=sort(kindofngrams, decreasing=TRUE)[1:100])
  })
  
  # quick summary for filtered subcorpus
  output$numDocumentsFilter <- reactivePrint(function() {
    dataset <- datasetInput() 
    subcorpus <- tm_filter(dataset, FUN = searchFullText, input$word)
   summary(subcorpus)
  })
  
  # calculate N-grams for filtered subset
  output$viewNgramFilter <- reactivePrint(function() {
    dataset <- datasetInput()
    subcorpus <- tm_filter(dataset, FUN = searchFullText, input$word)
    kindofngramsFilter <- textcnt(subcorpus, method = "string",n=input$numngrams)
    data.frame(count=sort(kindofngramsFilter, decreasing=TRUE)[1:100])
    
  })
  
  # calculate and plot correlation matrix
  output$corrPlot <- reactivePlot(function() {
    dtm <- dtmInput()  
    count <- input$count+1
    words <- names(findAssocs(dtm, input$word,input$prob)[2:count])
    oi <- as.matrix(dtm)
    find <- colnames(oi) %in% words
    corr <- cor(oi[,find]) 
    corrplot(corr)
  })
  
  # calculate and plot MDS
  output$mds <- reactivePlot(function() {
    dtm <- dtmInput()  
    count <- input$count+1
    words <- names(findAssocs(dtm, input$word,input$prob)[2:count])
    oi <- as.matrix(dtm)
    find <- colnames(oi) %in% words
    corr <- cor(oi[,find])
    d <- dist(corr) # euclidean distances between the rows
    fit <- cmdscale(d,eig=TRUE, k=2) # k is the number of dim
    x <- fit$points[,1]
    y <- fit$points[,2]
    plot(x, y, xlab="Coordinate 1", ylab="Coordinate 2", type="n")
    text(x, y, labels = row.names(corr))
  })
  
  # calculate and print summary for factory analysis
  output$viewFactor <- reactivePrint(function() {
    dtm <- dtmInput()  
    count <- input$count+1
    words <- names(findAssocs(dtm, input$word,input$prob)[2:count])
    oi <- as.matrix(dtm)
    find <- colnames(oi) %in% words
    corr <- cor(oi[,find])
    fit <- factanal(covmat=corr,factors=input$numfactors,rot="varimax")
    fit

  })
  
  # calculate and plot factory analysis
  output$factor <- reactivePlot(function() {
    dtm <- dtmInput()  
    count <- input$count+1
    words <- names(findAssocs(dtm, input$word,input$prob)[2:count])
    oi <- as.matrix(dtm)
    find <- colnames(oi) %in% words
    corr <- cor(oi[,find])
    fit <- factanal(covmat=corr,factors=input$numfactors,rot="varimax")
    load <- fit$loadings[,1:2] 
    plot(load,type="n") # set up plot 
    text(load,labels=row.names(corr))

  })
  
  # calculate and plot hierarchical clustering
  output$corrClust <- reactivePlot(function() {
    dtm <- dtmInput()  
    count <- input$count+1
    words <- names(findAssocs(dtm, input$word,input$prob)[2:count])
    oi <- as.matrix(dtm)
    find <- colnames(oi) %in% words
    d <- dist(cor(oi[,find]) , method = "euclidean")
    fit <- hclust(d, method="ward")
    plot(fit) 
  })




})