/**
* <summary> Constructor of VectorizedNerInstanceGenerator which takes input a {@link VectorizedDictionary}, a window size
* and a word format and sets corresponding attributes with these inputs.</summary>
* <param name="fsm">Morphological analyzer used to create multi-word expressions.</param>
* <param name="wordNet">WordNet for checking multii-word and single-word expressions.</param>
* <param name="dictionary">Dictionary in the vector form. Each word is stored in vector form in this dictionary.</param>
* <param name="windowSize">Number of previous (next) words to be considered in adding attributes.</param>
* <param name="format">Word vector format.</param>
*/
public VectorizedSemanticInstanceGenerator(FsmMorphologicalAnalyzer fsm, WordNet.WordNet wordNet,
VectorizedDictionary dictionary, int windowSize, WordFormat format) : base(fsm, wordNet)
{
this._format = format;
this._dictionary = dictionary;
this.windowSize = windowSize;
}
/**
* <summary> Constructor of VectorizedNerInstanceGenerator which takes input a {@link VectorizedDictionary}, a window size
* and a word format and sets corresponding attributes with these inputs.</summary>
* <param name="dictionary">Dictionary in the vector form. Each word is stored in vector form in this dictionary.</param>
* <param name="windowSize">Number of previous (next) words to be considered in adding attributes.</param>
* <param name="format">Word vector format.</param>
*/
public VectorizedShallowParseInstanceGenerator(VectorizedDictionary dictionary, int windowSize,
WordFormat format)
{
this._format = format;
this._dictionary = dictionary;
this.windowSize = windowSize;
}
/**
* <summary>Main method for training the Word2Vec algorithm. Depending on the training parameter, CBox or SkipGram algorithm
* is applied.</summary>
* <returns>Dictionary of word vectors.</returns>
*/
public VectorizedDictionary Train()
{
var result = new VectorizedDictionary(new TurkishWordComparator());
if (_parameter.IsCbow())
{
TrainCbow();
}
else
{
TrainSkipGram();
}
for (var i = 0; i < _vocabulary.Size(); i++)
{
result.AddWord(new VectorizedWord(_vocabulary.GetWord(i).GetName(), _wordVectors.GetRow(i)));
}
return(result);
}
