void mmEM::readFileXY(param myParam, string filename, vector_2str *wordX, vector_2str *wordY)
{
cout << "Reading file: " << filename << endl;
ifstream INPUTFILE;
INPUTFILE.open(filename.c_str());
if (! INPUTFILE)
{
cerr << "error: unable to open file " << filename << endl;
exit(-1);
}
while (! INPUTFILE.eof() )
{
string line;
vector<string> lineList;
// read each line and split column by space //
getline(INPUTFILE, line);
// ignore empty line
if (line == "")
{
continue;
}
// split by tab to get source and target //
if (myParam.inFormat == "l2p")
{
lineList = splitBySpace(line);
vector<string> t0,t1;
Tokenize(lineList[0], t0, "");
// lhuang
//Tokenize(lineList[1], t1, "");
wordX->push_back(t0);
wordY->push_back(splitBySpace(lineList[1]));
}
else if (myParam.inFormat == "news")
{
Tokenize(line, lineList, "\t");
// lhuang: no space on source (letters)
vector<string> t0;
Tokenize(lineList[0], t0, "");
wordX->push_back(t0);
wordY->push_back(splitBySpace(lineList[1]));
}
else
{
cerr << "ERROR: can't find input format type, plz. check --inFormat" << endl << endl;
}
if (lineList.size() != 2)
{
cerr << "Warning : missing either x or y word here, so skipped:" << endl << line << endl;
}
}
// close file //
INPUTFILE.close();
}
void readSentFile(const string &file, vector<vector<intern_string> > &sentences)
{
cerr << "Reading sentences from: " << file << "...";
ifstream TRAININ;
TRAININ.open(file.c_str());
if (! TRAININ)
{
cerr << "Error: can't read from file " << file << endl;
exit(-1);
}
string line;
while (getline(TRAININ, line))
{
vector<string> words;
splitBySpace(line, words);
vector<intern_string> intern_words(words.begin(), words.end());
sentences.push_back(intern_words);
if (sentences.size() % 1000000 == 0)
cerr << sentences.size() << "...";
}
cerr << endl;
TRAININ.close();
}
void m_dccchat::onUserCTCP(std::string server, std::string nick, std::string message) {
std::vector<std::string> messageParts = splitBySpace(message);
if (messageParts[0] == "DCC" && messageParts[1] == "CHAT" && messageParts[2] == "chat") {
if (activeConnections.find(server + "/" + nick) == activeConnections.end())
dccConnect(server, nick, messageParts[3], messageParts[4]);
else
sendNotice(server, nick, "You already have an active DCC chat session!");
}
}
void mmEM::readInitFile(param myParam)
{
// if there is an initial mapping file //
// read the initFile, fill up the limit set //
if (myParam.initFile != "")
{
cout << "Reading the initial file: " << myParam.initFile << endl;
ifstream INITFILE;
INITFILE.open(myParam.initFile.c_str());
if (! INITFILE)
{
cerr << "error: unable to open file " << myParam.initFile << endl;
exit(-1);
}
vector_initTable initCount;
while (! INITFILE.eof() )
{
string line;
vector<string> lineList;
getline(INITFILE, line);
initTable initTmp;
if (line == "")
{
continue;
}
//string t0,t1,t3;
//if (myParam.inFormat == "l2p")
//{
// should read it as the model format //
lineList = splitBySpace(line);
initTmp.xstring = lineList[0];
initTmp.ystring = lineList[1];
limitSet.insert(initTmp.xstring + "|" + initTmp.ystring);
if (lineList.size() > 2)
{
initTmp.prob = (long double)atof(lineList[2].c_str());
}
else
{
initTmp.prob = 1;
}
initCount.push_back(initTmp);
}
INITFILE.close();
}
}
void readWeightsFile(ifstream &TRAININ, vector<float> &weights) {
string line;
while (getline(TRAININ, line) && line != "")
{
vector<string> items;
splitBySpace(line, items);
if (items.size() != 1)
{
cerr << "Error: weights file should have only one weight per line" << endl;
exit(-1);
}
weights.push_back(boost::lexical_cast<float>(items[0]));
}
}
void readWordsFile(ifstream &TRAININ, vector<string> &word_list)
{
string line;
while (getline(TRAININ, line) && line != "")
{
vector<string> words;
splitBySpace(line, words);
if (words.size() != 1)
{
cerr << "Error: vocabulary file must have only one word per line" << endl;
exit(-1);
}
word_list.push_back(words[0]);
}
}
// Read a data file of unknown size into a flat vector<int>.
// If this takes too much memory, we should create a vector of minibatches.
void readSentFile(const string &filename,
vector<vector <int> > &data,
int minibatch_size,
data_size_t &num_tokens)
{
cerr << "Reading input sentences from file " << filename << ": ";
ifstream DATAIN(filename.c_str());
if (!DATAIN)
{
cerr << "Error: can't read data from file " << filename<< endl;
exit(-1);
}
vector<int> data_vector;
string line;
long long int n_lines = 0;
while (getline(DATAIN, line))
{
vector<string> ngram;
splitBySpace(line, ngram);
/*
if (ngram_size == 0)
ngram_size = ngram.size();
if (ngram.size() != ngram_size)
{
cerr << "Error: expected " << ngram_size << " fields in instance, found " << ngram.size() << endl;
exit(-1);
}
*/
vector<int> int_ngram;
for (int i=0;i<ngram.size();i++)
int_ngram.push_back(boost::lexical_cast<int>(ngram[i]));
data.push_back(int_ngram);
num_tokens += int_ngram.size();
n_lines++;
if (minibatch_size && n_lines % (minibatch_size * 10000) == 0)
cerr << n_lines/minibatch_size << "...";
}
cerr << "done." << endl;
DATAIN.close();
}
void model::readConfig(ifstream &config_file)
{
string line;
vector<string> fields;
int ngram_size, vocab_size, input_embedding_dimension, num_hidden, output_embedding_dimension;
activation_function_type activation_function = this->activation_function;
while (getline(config_file, line) && line != "")
{
splitBySpace(line, fields);
if (fields[0] == "ngram_size")
ngram_size = lexical_cast<int>(fields[1]);
else if (fields[0] == "vocab_size")
input_vocab_size = output_vocab_size = lexical_cast<int>(fields[1]);
else if (fields[0] == "input_vocab_size")
input_vocab_size = lexical_cast<int>(fields[1]);
else if (fields[0] == "output_vocab_size")
output_vocab_size = lexical_cast<int>(fields[1]);
else if (fields[0] == "input_embedding_dimension")
input_embedding_dimension = lexical_cast<int>(fields[1]);
else if (fields[0] == "num_hidden")
num_hidden = lexical_cast<int>(fields[1]);
else if (fields[0] == "output_embedding_dimension")
output_embedding_dimension = lexical_cast<int>(fields[1]);
else if (fields[0] == "activation_function")
activation_function = string_to_activation_function(fields[1]);
else if (fields[0] == "version")
{
int version = lexical_cast<int>(fields[1]);
if (version != 1)
{
cerr << "error: file format mismatch (expected 1, found " << version << ")" << endl;
exit(1);
}
}
else
cerr << "warning: unrecognized field in config: " << fields[0] << endl;
}
resize(ngram_size,
input_vocab_size,
output_vocab_size,
input_embedding_dimension,
num_hidden,
output_embedding_dimension);
set_activation_function(activation_function);
}
void mmEM::readAlignerFromFile(param myParam)
{
cout << "Reading aligner model from file : " << myParam.alignerIn << endl;
//clear model parameters
probs.clear();
counts.clear();
ifstream ALIGNERIN;
ALIGNERIN.open(myParam.alignerIn.c_str());
if (! ALIGNERIN)
{
cerr << "Error : can't read aligner model from file " << myParam.alignerIn << endl;
exit(-1);
}
while (! ALIGNERIN.eof())
{
string line;
vector<string> lineList;
getline(ALIGNERIN, line);
if (line == "")
{
continue;
}
lineList = splitBySpace(line);
if (lineList.size() != 3)
{
cerr << "Error : aligner model is in the wrong format " << endl << line << endl;
exit(-1);
}
// problem with long double when reading from string //
//probs[lineList[0]][lineList[1]] = convertTo<long double>(lineList[2]);
probs[lineList[0]][lineList[1]] = (long double)atof(lineList[2].c_str());
}
ALIGNERIN.close();
}
void allPhonemeSet::addFromFile(string filename, bool limitCandidate)
{
string line;
vector<string> lineList;
if (limitCandidate)
{
ifstream FILEIN;
FILEIN.open(filename.c_str());
while (! FILEIN.eof())
{
getline(FILEIN,line);
lineList = splitBySpace(line);
if (lineList.size() > 1)
{
addPhoneme(lineList[1],lineList[0],limitCandidate);
}
}
FILEIN.close();
}
}
void mmEM::initialization(param myParam, vector_2str stringX, vector_2str stringY)
{
if (myParam.limitPair)
{
readInitFile(myParam);
}
if (stringX.size() != stringY.size())
{
cerr << "error: data are not in pairs of x and y " << endl;
cerr << "# of x instances : " << stringX.size() << endl;
cerr << "# of y instances : " << stringY.size() << endl;
exit(-1);
}
// initialization with uniform distribution all possible alignments //
long double totalCount = 0; // keep track how many observations
// for each x and y pair //
for (int i=0; i < stringX.size(); i++)
{
// over lengths of x and y
for (int xl = 0; xl < stringX[i].size(); xl++)
{
for (int yl = 0; yl < stringY[i].size(); yl++)
{
if (myParam.delX)
{
for (int j=0; (j < myParam.maxX) && (xl-j >= 0); j++)
{
//string ssX = stringX[i].substr(xl-j,j+1);
string ssX = join(stringX[i], xl-j , j+1);
counts[ssX][myParam.nullChar] = 1;
}
}
if (myParam.delY)
{
for (int k=0; (k < myParam.maxY) && (yl-k >=0); k++)
{
// string ssY = stringY[i].substr(yl-k,k+1);
string ssY = join(stringY[i], yl-k, k+1);
counts[myParam.nullChar][ssY] = 1;
}
}
for (int j = 0; (j < myParam.maxX) && (xl-j >= 0); j++)
{
for (int k=0; (k < myParam.maxY) && (yl-k >=0); k++)
{
//string ssX = stringX[i].substr(xl-j,j+1);
//string ssY = stringY[i].substr(yl-k,k+1);
string ssX = join(stringX[i], xl-j, j+1);
string ssY = join(stringY[i], yl-k, k+1);
// if it defines a limit set //
if (myParam.limitPair)
{
if (limitSet.find(ssX + myParam.sepChar + ssY) == limitSet.end())
{
continue;
}
}
counts[ssX][ssY] = 1;
}
}
}
}
}
// if there is an initial mapping file //
// while reading the initFile, fill up the limit set //
if (myParam.initFile != "")
{
cout << "Reading the initial file: " << myParam.initFile << endl;
ifstream INITFILE;
INITFILE.open(myParam.initFile.c_str());
if (! INITFILE)
{
cerr << "error: unable to open file " << myParam.initFile << endl;
exit(-1);
}
vector_initTable initCount;
while (! INITFILE.eof() )
{
string line;
vector<string> lineList;
getline(INITFILE, line);
initTable initTmp;
if (line == "")
{
continue;
}
//string t0,t1,t3;
//if (myParam.inFormat == "l2p")
//{
// should read it as the model format //
lineList = splitBySpace(line);
initTmp.xstring = lineList[0];
initTmp.ystring = lineList[1];
if (lineList.size() > 2)
{
initTmp.prob = (long double)atof(lineList[2].c_str());
}
else
{
initTmp.prob = 1;
}
initCount.push_back(initTmp);
}
INITFILE.close();
// sort initCount //
cout << "Sorting the initial count table" << endl;
sort(initCount.begin(), initCount.end(), initTableSortedFn);
long double total_add_prob = 0;
for (vector_initTable::iterator pos = initCount.begin(); pos != initCount.end(); pos++)
{
if ((total_add_prob < myParam.initProbCut) || (pos->prob == 1))
{
counts[pos->xstring][pos->ystring] += ((counts[pos->xstring].size() * 10) + stringX.size()) * pos->prob ;
}
else
{
break;
}
if (pos->prob != 1)
{
total_add_prob += pos->prob;
}
}
}
}
