Learning to Classify Text

Target concept 136#136

1. Represent each document by vector of words
   • one attribute per word position in document
2. Learning: Use training examples to estimate
   • 137#137
   • 138#138
   • 139#139
   • 140#140

Naive Bayes conditional independence assumption

141#141

where 142#142 is probability that word in position 143#143 is 144#144, given 112#112

one more assumption: 145#145

LEARN/SMALL>_NAIVE/SMALL>_BAYES/SMALL>_TEXT(146#146)

1. collect all words and other tokens that occur in 147#147

 $$

148#148 all distinct words and other tokens in 147#147

2. calculate the required 149#149 and 150#150 probability terms

 $$

For each target value 151#151 in 152#152 do

• 153#153 subset of 147#147 for which the target value is 151#151

• 154#154

• 155#155 a single document created by concatenating all members of 156#156

• 157#157 total number of words in 158#158 (counting duplicate words multiple times)

• for each word 159#159 in 160#160
  • 161#161 number of times word 159#159 occurs in 158#158

  • 162#162

CLASSIFY/SMALL>_NAIVE/SMALL>_BAYES/SMALL>_TEXT(163#163)

• 164#164 all word positions in 163#163 that contain tokens found in 160#160

• Return 165#165, where
  
  166#166