private float scorePhrase(final TargetPhrase targetPhrase) {
float score = 0;
for (int targetIndex = 0; targetIndex < targetPhrase.getSize(); targetIndex++) {
float sum = 0;
final Word targetWord = targetPhrase.getWord(targetIndex);
VERBOSE.v(2, "glm " + targetWord + ": ");
Map<Word, Float> dh = m_hash.get(targetWord);
if (dh != null) {
Float f = dh.get(m_bias);
if (f != null) {
VERBOSE.v(2, "*BIAS* " + f);
sum += f;
}
Set<Word> alreadyScored = new HashSet<Word>();
for (int inputIndex = 0; inputIndex < m_input.getSize(); inputIndex++) {
final Word inputWord = m_input.getWord(inputIndex);
if (!alreadyScored.contains(inputWord)) {
Float w = dh.get(inputWord);
if (w != null) {
VERBOSE.v(2, " " + inputWord + " " + w);
sum += w;
}
alreadyScored.add(inputWord);
}
}
}
final Map<Word, Float> targetWordHash = m_hash.get(targetWord);
if (targetWordHash != null) {
Float inputWordHash = targetWordHash.get(m_bias);
if (inputWordHash != null) {
VERBOSE.v(2, "*BIAS* " + inputWordHash);
sum += inputWordHash;
}
Set<Word> alreadyScored = new HashSet<Word>(); // do not score a
// word twice
for (int inputIndex = 0; inputIndex < m_input.getSize(); inputIndex++) {
final Word inputWord = m_input.getWord(inputIndex);
if (!alreadyScored.contains(inputWord)) {
Float inputWordHash0 = targetWordHash.get(inputWord);
if (inputWordHash0 != null) {
VERBOSE.v(2, " " + inputWord + " " + inputWordHash0);
sum += inputWordHash0;
}
alreadyScored.add(inputWord);
}
}
}
// Hal Daume says: 1/( 1 + exp [ - sum_i w_i * f_i ] )
VERBOSE.v(2, " p=" + Util.floorScore((float) Math.log(1 / (1 + Math.exp(-sum)))) + '\n');
score += Util.floorScore((float) Math.log(1 / (1 + Math.exp(-sum))));
}
return score;
}
public void sendAuth() throws Exception {
String username = "******";
String password = "******";
int passwordHash[] = DoubleHash.doubleHash(password.toLowerCase(), clientToken, serverToken);
BNetProtocolPacket p = new BNetProtocolPacket(BNetProtocolPacketId.SID_LOGONRESPONSE2);
p.writeDWord(clientToken);
p.writeDWord(serverToken);
p.writeDWord(passwordHash[0]);
p.writeDWord(passwordHash[1]);
p.writeDWord(passwordHash[2]);
p.writeDWord(passwordHash[3]);
p.writeDWord(passwordHash[4]);
p.writeNTString(username);
p.sendPacket(BNetOutputStream);
}
private void loadData(String filePath, final int[] inFactors, final int[] outFactors)
throws IOException {
FactorCollection factorCollection = FactorCollection.instance();
final String factorDelimiter = StaticData.instance().getFactorDelimiter();
VERBOSE.v(2, "Loading global lexical model from file " + filePath + "\n");
// m_inputFactors = new FactorMask(inFactors);
// m_outputFactors = new FactorMask(outFactors);
BufferedReader inFile = new BufferedReader(new FileReader(filePath));
// reading in data one line at a time
int lineNum = 0;
String line;
while ((line = inFile.readLine()) != null) {
++lineNum;
String token[] = Util.tokenize(line, " ");
if (token.length != 3) // format checking
{
StringBuilder errorMessage = new StringBuilder();
errorMessage
.append("Syntax error at ")
.append(filePath)
.append(":")
.append(lineNum)
.append('\n')
.append(line)
.append('\n');
UserMessage.add(errorMessage.toString());
System.exit(0);
}
// create the output word
Word outWord = new Word();
String[] factorString = Util.tokenize(token[0], factorDelimiter);
for (int i = 0; i < outFactors.length; i++) {
final FactorDirection direction = FactorDirection.Output;
final int factorType = outFactors[i];
final Factor factor = factorCollection.addFactor(direction, factorType, factorString[i]);
outWord.setFactor(factorType, factor);
}
// create the input word
Word inWord = new Word();
factorString = Util.tokenize(token[1], factorDelimiter);
for (int i = 0; i < inFactors.length; i++) {
final FactorDirection direction = FactorDirection.Input;
final int factorType = inFactors[i];
final Factor factor = factorCollection.addFactor(direction, factorType, factorString[i]);
inWord.setFactor(factorType, factor);
}
// maximum entropy feature score
float score = Float.valueOf(token[2]);
// std::cerr << "storing word " << *outWord << " " << *inWord <<
// " " << score << endl;
// store feature in hash
Map<Word, Float> dh = m_hash.get(outWord);
if (dh == null) {
dh = new HashMap<Word, Float>();
dh.put(inWord, score);
m_hash.put(outWord, dh);
} else {
dh.put(inWord, score);
outWord = null;
}
}
}
