/// <summary>
/// Acrescenta as observações ao conjunto de treinamento
/// e adapta o modelo.
/// </summary>
public void Train(List <Feature> observations)
{
var o = quantizer.Quantize(observations);
if (o != null)
{
trainingSet.Add(o);
if (Parameterization.IncrementalTraining || trainingSet.Count == Parameterization.TrainingSetSize)
{
Hmm.Learn(trainingSet.ToArray());
}
}
}
/**
* <summary>Train method for the Hmm pos tagger. The algorithm trains an Hmm from the corpus, where corpus constitutes
* as an observation array.</summary>
*
* <param name="corpus">Training data for the tagger.</param>
*/
public void Train(PosTaggedCorpus corpus)
{
var emittedSymbols = new List <string> [corpus.SentenceCount()];
for (var i = 0; i < emittedSymbols.Length; i++)
{
emittedSymbols[i] = new List <string>();
for (var j = 0; j < corpus.GetSentence(i).WordCount(); j++)
{
var word = (PosTaggedWord)corpus.GetSentence(i).GetWord(j);
emittedSymbols[i].Add(word.GetTag());
}
}
_hmm = new Hmm1 <string, Word>(corpus.GetTagList(), emittedSymbols, corpus.GetAllWordsAsArray());
}
/// <summary>
/// Devolve a probabilidade das observações terem sido geradas por esse modelo.
/// </summary>
public double Evaluate(List <Feature> observations)
{
var q = quantizer.Quantize(observations);
return(Hmm.Evaluate(q));
}
