From 4ecf08a5f4ee347b2ee5b985618205ef07df2970 Mon Sep 17 00:00:00 2001
From: EeshanChatterjee <eeshanchatterjee@gmail.com>
Date: Tue, 7 Jan 2014 12:22:26 +0530
Subject: [PATCH 1/4] Update sentiment_analysis.R

---
 sentiment_analysis.R | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/sentiment_analysis.R b/sentiment_analysis.R
index fc33911..4503b1b 100644
--- a/sentiment_analysis.R
+++ b/sentiment_analysis.R
@@ -74,4 +74,8 @@ 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)
\ No newline at end of file
+binom.test(confTable[1,1] + confTable[2,2], nrow(results), p=0.5)
+
+#  chi sqr polarity test of the confusion table
+
+# PMI-based polarity test for confTable

From 423ee2f05df91087d2fc7310788f88201857bfa4 Mon Sep 17 00:00:00 2001
From: EeshanChatterjee <eeshan.chatterjee@mu-sigma.com>
Date: Tue, 7 Jan 2014 12:27:36 +0530
Subject: [PATCH 2/4] added polarity funcs

---
 polarityFuncs.R | 113 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 113 insertions(+)
 create mode 100644 polarityFuncs.R

diff --git a/polarityFuncs.R b/polarityFuncs.R
new file mode 100644
index 0000000..54a419b
--- /dev/null
+++ b/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)
+}

From e838b41980f35726b40aabe69aa78b20b3fc33fc Mon Sep 17 00:00:00 2001
From: EeshanChatterjee <eeshanchatterjee@gmail.com>
Date: Tue, 7 Jan 2014 12:31:22 +0530
Subject: [PATCH 3/4] Update sentiment_analysis.R

calling the polarity functions
---
 sentiment_analysis.R | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/sentiment_analysis.R b/sentiment_analysis.R
index 4503b1b..3f673b8 100644
--- a/sentiment_analysis.R
+++ b/sentiment_analysis.R
@@ -76,6 +76,11 @@ confTable
 #run a binomial test for confidence interval of results
 binom.test(confTable[1,1] + confTable[2,2], nrow(results), p=0.5)
 
-#  chi sqr polarity test of the confusion table
+source("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

From 14f8d5496716a99266d5678322b330d4524a8909 Mon Sep 17 00:00:00 2001
From: EeshanChatterjee <eeshanchatterjee@gmail.com>
Date: Tue, 7 Jan 2014 12:32:02 +0530
Subject: [PATCH 4/4] Update sentiment_analysis.R

---
 sentiment_analysis.R | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sentiment_analysis.R b/sentiment_analysis.R
index 3f673b8..5d8347d 100644
--- a/sentiment_analysis.R
+++ b/sentiment_analysis.R
@@ -76,7 +76,7 @@ confTable
 #run a binomial test for confidence interval of results
 binom.test(confTable[1,1] + confTable[2,2], nrow(results), p=0.5)
 
-source("polarityFuncs.R")
+source("~/Documents/GitHub/polarityFuncs.R")
 
 #  chi sqr polarity test of the confusion table
 chiSqrPV = calcChiSqrPV(as.vector(confTable))