Added 2 polarity conparison fiunctions for the confusion table

polarityFuncs.R

@@ -0,0 +1,113 @@
+# ======================================================================================#
+# This Script contains functions designed to return a 
+# numerical value indicating the strength of polarity, or the polatiry value
+# of a word or phrase.
+
+# The input taken by the functions are in the form of 2x2 Contingency matrix, eg.
+#                              pos          neg
+#                   c   |    f(c,pos)    f(c,neg)
+#                   ~c  |   f(~c,pos)   f(~c,neg)
+# ======================================================================================#
+
+
+# ================================================================================
+# Function: calcChiSqrPV
+# Task: Calculate the Chi Square based Polarity Value from a contingency matrix
+# Input: Numeric vector/Matrix representing the contingeny matrix
+# Output: Polarity Value
+# ================================================================================
+
+calcChiSqrPV<-function(conMat){
+
+  # Chi Square based polarity value
+  mat<- matrix(as.numeric(conMat), ncol=2, byrow=TRUE)
+  expmat<- matrix(data=NA, nrow=2, ncol=2)
+
+  # Sum of all elements of contingency matrix
+  sum<-0
+
+  #Loop to get the sum
+  for(i in 1:2){ # row
+    for(j in 1:2){ #coloumn
+      sum<-sum+mat[i,j]
+    }
+  }
+
+  # For-loop to genetare matrix of expected values
+  for(i in 1:2){ # row
+    for(j in 1:2){ #coloumn
+      rowsum<-0
+      colsum<-0
+
+      for(k in 1:2){
+        rowsum<- rowsum+mat[i,k]
+      }
+
+      for(k in 1:2){
+        colsum<- colsum+mat[k,j]
+      }
+
+      expmat[i,j]<- (rowsum*colsum)/sum
+    }
+  }
+
+  chisqr<-0
+
+  # Loop to calculate the chi-square value
+  for(i in 1:2){ # row
+    for(j in 1:2){ #coloumn
+
+      chisqr<- chisqr+ ((mat[i,j]-expmat[i,j])^2)/expmat[i,j]
+
+    }
+  }
+
+  Pc_pos<-mat[1,1]/(mat[1,1]+mat[2,1])
+  Pc_neg<-mat[1,2]/(mat[1,2]+mat[2,2])
+
+  if(Pc_pos<Pc_neg){
+    chisqr<- chisqr*-1
+  }
+
+  return(chisqr)
+
+}
+
+
+# ================================================================================
+# Function: calcPmiPV
+# Task: Calculate the PMI based Polarity Value from a contingency matrix
+# Input: Numeric vector/Matrix representing the contingeny matrix
+# Output: Polarity Value
+#=================================================================================
+
+calcPmiPV<- function(conMat){
+
+  # Pointwise Mutual Information
+  # Contingency Table
+  mat<- matrix(as.numeric(conMat), ncol=2, byrow=TRUE)
+
+  # cpos, cneg, nocpos, nocneg
+  cpos<-mat[1,1]
+  cneg<-mat[1,2]
+  nocpos<-mat[2,1]
+  nocneg<-mat[2,2]
+
+  sum<-cpos+cneg+nocpos+nocneg
+
+  # 
+  Pcpos<- cpos/sum
+  Pcneg<- cneg/sum
+  Ppos<- cpos+nocpos/sum
+  Pneg<- cneg+nocneg/sum
+  Pc<- cpos+cneg/sum
+
+  # PMI for cpos & cneg
+  PMIcpos<- log2(Pcpos/(Pc*Ppos))
+  PMIcneg<- log2(Pcneg/(Pc*Pneg))
+
+  # Polarity Value (PV) is PMI(c,pos) - PMI(c,neg)
+  PV<- (PMIcpos-PMIcneg)
+
+  return(PV)
+}

sentiment_analysis.R

@@ -74,4 +74,13 @@ confTable <- table(predict(classifier, results), results[,6], dnn=list('predicte
 confTable
 
 #run a binomial test for confidence interval of results
-binom.test(confTable[1,1] + confTable[2,2], nrow(results), p=0.5)
+binom.test(confTable[1,1] + confTable[2,2], nrow(results), p=0.5)
+
+source("~/Documents/GitHub/polarityFuncs.R")
+
+#  chi sqr polarity test of the confusion table
+chiSqrPV = calcChiSqrPV(as.vector(confTable))
+chiSqrPV
+# PMI-based polarity test for confTable
+PmiPV = calcPmiPV(as.vector(confTable))
+PmiPV