public static double[][] extractBigrams(
File file, MapSymbolTable symbolTable, TokenizerFactory tokenizerFactory, String charset)
throws Exception {
char[] cs = Files.readCharsFromFile(file, charset);
String[] tokens = tokenizerFactory.tokenizer(cs, 0, cs.length).tokenize();
System.out.println(" Number of tokens=" + tokens.length);
int[] symbols = new int[tokens.length];
for (int i = 0; i < tokens.length; ++i) {
symbols[i] =
Strings.allLetters(tokens[i].toCharArray()) ? symbolTable.getOrAddSymbol(tokens[i]) : -1;
}
int numSymbols = symbolTable.numSymbols();
System.out.println(" Number of distinct tokens=" + numSymbols);
System.out.println(" #Matrix entries=" + numSymbols * numSymbols);
double[][] values = new double[numSymbols][numSymbols];
for (int i = 0; i < numSymbols; ++i) Arrays.fill(values[i], 0.0);
for (int i = 1; i < symbols.length; ++i) {
int left = symbols[i - 1];
int right = symbols[i];
if (left >= 0 && right >= 0) values[symbols[i - 1]][symbols[i]] += 1.0;
}
return values;
}
public static void main(String[] args) throws ClassNotFoundException, IOException {
if (args.length != 3) {
System.out.println("No Arguments. No Model Trainer.");
return;
}
TESTING_DIR = new File(args[0]);
TESTING_RESULTS_DIR = new File(args[1]);
MODELS_DIR = new File(args[2]);
CATEGORIES = TESTING_DIR.list();
NGRAM_SIZE = CATEGORIES.length;
File[] models = MODELS_DIR.listFiles();
PartsOfSpeech = new String[models.length];
compiledClassifiers = new HashMap<String, LMClassifier>();
for (int i = 0; i < models.length; i++) {
LMClassifier classifier = loadModel(models[i].getAbsolutePath());
PartsOfSpeech[i] = models[i].getName().substring(0, models[i].getName().length() - 6);
compiledClassifiers.put(PartsOfSpeech[i], classifier);
}
standardClassifier = compiledClassifiers.get("all");
// testing
// ConfusionMatrix confMatrix = new ConfusionMatrix(CATEGORIES);
ClassifierEvaluator evaluator = new ClassifierEvaluator(standardClassifier, CATEGORIES);
int hits = 0, misses = 0, hits_standard = 0, misses_standard = 0;
for (int i = 0; i < CATEGORIES.length; i++) {
File classDir = new File(TESTING_DIR, CATEGORIES[i]);
String[] testingFiles = classDir.list();
for (int j = 0; j < testingFiles.length; j++) {
// String text = Files.readFromFile(new File(classDir,testingFiles[j]));
System.out.print("Testing on " + CATEGORIES[i] + "/" + testingFiles[j] + " ");
// evaluator.addCase(CATEGORIES[i],text);
Document d = new Document(new File(classDir, testingFiles[j]));
Set<Document> s = new HashSet<Document>();
s.add(d);
POSFinder p = new POSFinder(s);
p.process();
JointClassification jc = jointJointClassify(d);
String bestCategory = jc.bestCategory();
JointClassification jc_standard = standardClassifier.classify(d.mText_str);
// evaluator.addClassification(CATEGORIES[i], new
// com.aliasi.classify.Classification(bestCategory));
evaluator.addCase(CATEGORIES[i], d.mText_str);
System.out.println("Got best category of: " + bestCategory);
System.out.println("(standard) Got best category of: " + jc_standard.bestCategory());
System.out.println("---------------");
File bestCategoryDirectory = new File(TESTING_RESULTS_DIR, bestCategory);
File bestCategoryDirectory_inTest = new File(TESTING_DIR, bestCategory);
File bestCategoryDirectory_inTest_standard =
new File(TESTING_DIR, jc_standard.bestCategory());
if (!bestCategoryDirectory.exists()) bestCategoryDirectory.mkdir();
boolean hit = false;
for (String ss : bestCategoryDirectory_inTest.list())
if (ss.equals(d.mFile.getName())) hit = true;
if (hit) {
hits++;
System.err.println(
"-------------HIT! P()="
+ (int) (jc.conditionalProbability(0) * 100)
+ "% ("
+ hits
+ " hits)------------");
} else {
misses++;
System.err.println(
"-------------MISS! P()="
+ (int) (jc.conditionalProbability(0) * 100)
+ "% ("
+ misses
+ " misses)------------");
}
hit = false;
for (String ss : bestCategoryDirectory_inTest_standard.list())
if (ss.equals(d.mFile.getName())) hit = true;
if (hit) {
hits_standard++;
System.err.println(
"-----(standard)---HIT! P()="
+ (int) (jc_standard.conditionalProbability(0) * 100)
+ "% ("
+ hits_standard
+ " hits)------------");
} else {
misses_standard++;
System.err.println(
"----(standard)---MISS! P()="
+ (int) (jc_standard.conditionalProbability(0) * 100)
+ "% ("
+ misses_standard
+ " misses)------------");
}
if (hits_standard < hits)
System.err.println(
"Akshat's Classifier is WINNING <"
+ ((double) hits / (double) (hits + misses))
+ "> to <"
+ (double)
((double) hits_standard / ((double) hits_standard + (double) misses_standard))
+ ">");
else
System.err.println(
"Akshat's Classifier is LOSING <"
+ ((double) hits / (double) (hits + misses))
+ "> to <"
+ (double)
((double) hits_standard / ((double) hits_standard + (double) misses_standard))
+ ">");
// DocTreeWidget.copyFile(d.mFile, new File(bestCategoryDirectory,d.mFile.getName()));
// //this is just a file copy function
}
}
ConfusionMatrix confMatrix = evaluator.confusionMatrix();
String myresults =
"Hits: "
+ hits
+ ", Misses: "
+ misses
+ ", Total Accuracy: "
+ ((double) hits / (double) (hits + misses));
myresults +=
"\n(standard) Hits :"
+ hits_standard
+ ", Misses: "
+ misses_standard
+ " Total accuracy = "
+ (double)
((double) hits_standard / ((double) hits_standard + (double) misses_standard));
System.out.println(myresults);
com.aliasi.util.Files.writeStringToFile(
myresults, new File(TESTING_RESULTS_DIR, "results.txt"));
}