private void extractVocabulary(String directoryPath, int classID) throws IOException {
File dir = new File(directoryPath);
File[] directoryListing = dir.listFiles();
numDocs += directoryListing.length;
numDocsOfEachClass[classID] = directoryListing.length;
TreeMap<String, Integer> thisClassVocab = new TreeMap<String, Integer>();
if (directoryListing != null) {
for (File child : directoryListing) {
System.out.print(".");
FileInputStream fis = new FileInputStream(child);
BufferedReader br = new BufferedReader(new InputStreamReader(fis));
String wholeFile = "";
String line = "";
while (line != null) {
line = br.readLine();
wholeFile += line + " ";
}
for (String term : Utils.tokenizeString(wholeFile)) {
if (!term.matches(".*\\d+.*")) {
if (!vocabulary.contains(term)) vocabulary.add(term);
if (!thisClassVocab.containsKey(term)) {
thisClassVocab.put(term, 1);
} else if (thisClassVocab.containsKey(term)) {
int temp = thisClassVocab.get(term) + 1;
thisClassVocab.put(term, new Integer(temp));
}
}
}
br.close();
}
}
vocabOfClasses.add(thisClassVocab);
}
public String apply(String newDocPath) throws IOException {
FileInputStream fis = new FileInputStream(new File(newDocPath));
BufferedReader br = new BufferedReader(new InputStreamReader(fis));
String newDocContent = "";
String line = "";
while (line != null) {
line = br.readLine();
newDocContent += line + " ";
}
double[] score = new double[numClasses];
List<String> newDocTokens = Utils.tokenizeString(newDocContent);
TreeMap<String, Integer> termFreq = new TreeMap<String, Integer>();
for (String str : newDocTokens) {
if (vocabulary.contains(str)) {
if (!termFreq.containsKey(str)) termFreq.put(str, 1);
else {
int temp = termFreq.get(str) + 1;
termFreq.put(str, temp);
}
}
}
br.close();
for (int c = 0; c < numClasses; c++) {
score[c] = Math.log10(prior[c]);
for (String t : termFreq.keySet()) {
score[c] += Math.log10(condprob[vocabulary.indexOf(t)][c]);
}
}
// return class id
return classNames[Utils.argmax(score)];
}