/**
* /// Phoneticize a word
* ///
* /// @param entry
* /// the word to phoneticize transformed to an ArrayList of Strings
* /// (each element hold a single character)
* /// @param nbest
* /// the number of distinct pronunciations to return
* /// @return the pronunciation(s) of the input word
*/
public List <Path> Phoneticize(List <String> entry, int nbest)
{
var efst = EntryToFsa(entry);
var s = efst.Semiring;
Compose.Augment(1, efst, s);
ArcSort.Apply(efst, new OLabelCompare());
var result = Compose.compose(efst, _epsilonFilter, s, true);
ArcSort.Apply(result, new OLabelCompare());
result = Compose.compose(result, _g2Pmodel, s, true);
Project.Apply(result, ProjectType.Output);
if (nbest == 1)
{
result = NShortestPaths.Get(result, 1, false);
}
else
{
// Requesting 10 times more best paths than what was asking
// as there might be several paths resolving to same pronunciation
// due to epsilon transitions.
// I really hate cosmological constants :)
result = NShortestPaths.Get(result, nbest * 10, false);
}
// result = NShortestPaths.get(result, nbest, false);
result = RmEpsilon.Get(result);
var paths = FindAllPaths(result, nbest, _skipSeqs, Tie);
return(paths);
}
///**
///// Create a decoder by loading the serialized model from a specified
///// filename
/////
///// @param g2pmodel_file
///// the filename of the serialized model
// */
//public G2PConverter(String g2pmodel_file)
//{
// g2pmodel = ImmutableFst.loadModel(g2pmodel_file);
// init();
//}
/**
* /// Initialize the decoder
*/
private void Init()
{
_skipSeqs.Add(Eps);
_skipSeqs.Add(Sb);
_skipSeqs.Add(Se);
_skipSeqs.Add(Skip);
_skipSeqs.Add("-");
// keep an augmented copy (for compose)
Compose.Augment(0, _g2Pmodel, _g2Pmodel.Semiring);
ArcSort.Apply(_g2Pmodel, new ILabelCompare());
var isyms = _g2Pmodel.Isyms;
LoadClusters(isyms);
// get epsilon filter for composition
_epsilonFilter = Compose.GetFilter(_g2Pmodel.Isyms, _g2Pmodel.Semiring);
ArcSort.Apply(_epsilonFilter, new ILabelCompare());
}