//Called when we start translating a new sentence
void PhraseDictionaryFeature::InitDictionary(const TranslationSystem* system, const InputType& source)
{
PhraseDictionary* dict;
if (m_useThreadSafePhraseDictionary) {
//thread safe dictionary should already be loaded
dict = m_threadSafePhraseDictionary.get();
} else {
//thread-unsafe dictionary may need to be loaded if this is a new thread.
if (!m_threadUnsafePhraseDictionary.get()) {
m_threadUnsafePhraseDictionary.reset(LoadPhraseTable(system));
}
dict = m_threadUnsafePhraseDictionary.get();
}
CHECK(dict);
dict->InitializeForInput(source);
}
int main(int argc, char** argv)
{
bool help;
string input_file;
string config_file;
po::options_description desc("Allowed options");
desc.add_options()
("help,h", po::value(&help)->zero_tokens()->default_value(false), "Print this help message and exit")
("input-file,i", po::value<string>(&input_file), "Input file")
("config-file,f", po::value<string>(&config_file), "Config file")
;
po::options_description cmdline_options;
cmdline_options.add(desc);
po::variables_map vm;
po::parsed_options parsed = po::command_line_parser(argc,argv).
options(cmdline_options).allow_unregistered().run();
po::store(parsed, vm);
po::notify(vm);
if (help) {
usage(desc, argv);
exit(0);
}
if (input_file.empty()) {
cerr << "ERROR: Please specify an input file" << endl << endl;
usage(desc, argv);
exit(1);
}
if (config_file.empty()) {
cerr << "ERROR: Please specify a config file" << endl << endl;
usage(desc, argv);
exit(1);
}
vector<string> mosesargs;
mosesargs.push_back(argv[0]);
mosesargs.push_back("-f");
mosesargs.push_back(config_file);
for (size_t i = 0; i < parsed.options.size(); ++i) {
if (parsed.options[i].position_key == -1 && !parsed.options[i].unregistered) continue;
/*
const string& key = parsed.options[i].string_key;
if (!key.empty()) {
mosesargs.push_back(key);
}
for (size_t j = 0; j < parsed.options[i].value.size(); ++j) {
const string& value = parsed.options[i].value[j];
if (!value.empty()) {
mosesargs.push_back(value);
}
}*/
for (size_t j = 0; j < parsed.options[i].original_tokens.size(); ++j) {
mosesargs.push_back(parsed.options[i].original_tokens[j]);
}
}
boost::scoped_ptr<Parameter> params(new Parameter());
char** mosesargv = new char*[mosesargs.size()];
for (size_t i = 0; i < mosesargs.size(); ++i) {
mosesargv[i] = new char[mosesargs[i].length() + 1];
strcpy(mosesargv[i], mosesargs[i].c_str());
}
if (!params->LoadParam(mosesargs.size(), mosesargv)) {
params->Explain();
exit(1);
}
if (!StaticData::LoadDataStatic(params.get(),argv[0])) {
exit(1);
}
const StaticData &staticData = StaticData::Instance();
const std::vector<FactorType> & input = staticData.GetInputFactorOrder();
//Find the phrase table to evaluate with
PhraseDictionary* phraseTable = NULL;
const vector<FeatureFunction*>& ffs = FeatureFunction::GetFeatureFunctions();
for (size_t i = 0; i < ffs.size(); ++i) {
PhraseDictionary* maybePhraseTable = dynamic_cast< PhraseDictionary*>(ffs[i]);
if (maybePhraseTable) {
UTIL_THROW_IF(phraseTable,util::Exception,"Can only score translations with one phrase table");
phraseTable = maybePhraseTable;
}
}
UTIL_THROW_IF(!phraseTable,util::Exception,"Unable to find scoring phrase table");
Sentence sentence;
phraseTable->InitializeForInput(sentence);
//
//Load and prune the phrase table. This is taken (with mods) from moses/TranslationModel/RuleTable/LoaderStandard.cpp
//
string lineOrig;
std::ostream *progress = NULL;
IFVERBOSE(1) progress = &std::cerr;
util::FilePiece in(input_file.c_str(), progress);
// reused variables
vector<float> scoreVector;
StringPiece line;
double_conversion::StringToDoubleConverter converter(double_conversion::StringToDoubleConverter::NO_FLAGS, NAN, NAN, "inf", "nan");
StringPiece previous;
while(true) {
try {
line = in.ReadLine();
} catch (const util::EndOfFileException &e) {
break;
}
util::TokenIter<util::MultiCharacter> pipes(line, "|||");
StringPiece sourcePhraseString(*pipes);
if (sourcePhraseString != previous) {
outputTopN(previous, phraseTable, input, cout);
previous = sourcePhraseString;
}
}
outputTopN(previous, phraseTable, input, cout);
return 0;
}
