//Extract word embedding features from current context
public VectorBase ExtractDenseFeature(int currentState, int numStates, List <string[]> features)
{
var fc = featureContext;
if (fc.ContainsKey(WORDEMBEDDING_CONTEXT))
{
var v = fc[WORDEMBEDDING_CONTEXT];
if (v.Count == 1)
{
var strKey = features[TruncPosition(currentState + v[0], 0, numStates)][preTrainedModelColumn];
return(preTrainedModel.GetTermVector(strKey));
}
var dense = new CombinedVector();
for (var j = 0; j < v.Count; j++)
{
var offset = currentState + v[j];
if (offset >= 0 && offset < numStates)
{
var strKey = features[offset][preTrainedModelColumn];
dense.Append(preTrainedModel.GetTermVector(strKey));
}
else
{
dense.Append(preTrainedModel.m_UnkEmbedding);
}
}
return(dense);
}
return(new SingleVector());
}
//Extract word embedding features from current context
public VectorBase ExtractDenseFeature(int currentState, int numStates, List <string[]> features)
{
var fc = m_FeatureConfiguration;
if (fc.ContainsKey(WORDEMBEDDING_CONTEXT) == true)
{
List <int> v = fc[WORDEMBEDDING_CONTEXT];
if (v.Count == 1)
{
string strKey = features[TruncPosition((int)currentState + v[0], 0, (int)numStates)][m_WordEmbeddingCloumn];
return(m_WordEmbedding.GetTermVector(strKey));
}
CombinedVector dense = new CombinedVector();
for (int j = 0; j < v.Count; j++)
{
int offset = currentState + v[j];
if (offset >= 0 && offset < numStates)
{
string strKey = features[offset][m_WordEmbeddingCloumn];
dense.Append(m_WordEmbedding.GetTermVector(strKey));
}
else
{
dense.Append(m_WordEmbedding.m_UnkEmbedding);
}
}
return(dense);
}
return(new SingleVector());
}