public DocumentViewer(Document documnet)
{
InitializeComponent();
v_TextBlock_Title.Text = documnet.Title;
v_TextBlock_OriginalText.Text = documnet.OriginalText;
v_TextBlock_ProcessedText.Text = documnet.PostProcessedText;
v_TextBlock_Keywords.Text = string.Concat(documnet.BagOfWords.Where(s => s.Value > 0)
.OrderByDescending(s => s.Value).Select(s => s.Key + ", "));
}
private void RecalculateDocuments()
{
//Calculating TF/IDF
#region TF IDF
foreach (var doc in _rawDocuments)
{
var docRep = new Document();
docRep.Title = Regex.Split(doc, "\r\n").FirstOrDefault();
docRep.OriginalText = string.Concat(Regex.Split(doc, "\r\n").Skip(1));
docRep.PostProcessedText = string.Concat(Regex.Replace(doc.ToLower(), @"[^\w\s]", "").Split(null).Select(s => _ps.stemTerm(s) + " "));
docRep.BagOfWords = new Dictionary<string, int>();
foreach (var keyword in _rawKeywords)
{
var temp = _ps.stemTerm(keyword);
docRep.BagOfWords[keyword] = docRep.PostProcessedText.Split(null).Count(s => s.Equals(_ps.stemTerm(temp)));
}
docRep.TF = new Dictionary<string, double>();
var tfMax = docRep.BagOfWords.Max(w => w.Value);
foreach (var word in docRep.BagOfWords)
{
if (tfMax <= 0)
{
docRep.TF[word.Key] = 0;
}
else
{
docRep.TF[word.Key] = (double)docRep.BagOfWords[word.Key] / tfMax;
}
}
_documents.Add(docRep);
}
//calculationg IDFs
foreach (var keyword in _rawKeywords)
{
var count = (double)_documents.Count;
var contains = _documents.Count(d => d.BagOfWords.Any(w => w.Key.Equals(keyword) && w.Value > 0));
IDF[keyword] = contains > 0 ? Math.Log10(count / contains) : 0;
}
//calculationg Vector Length
foreach (var docRep in _documents)
{
docRep.TFIDFVectorValue = DistanceN(docRep.TF.Select(tf => IDF[tf.Key] * tf.Value));
}
#endregion
//words Correlation
#region Correlation
foreach (var doc in _rawDocuments)
{
var wordsInDoc = GetWords(doc);
foreach (var word in wordsInDoc)
{
if (!allWords.Contains(word))
{
allWords.Add(word);
}
}
}
double[,] docsWordsArray = new double[_rawDocuments.Count, allWords.Count];
for (int i = 0; i < _rawDocuments.Count; i++)
{
var wordsInDoc = GetWords(_rawDocuments[i]);
for (int j = 0; j < allWords.Count; j++)
{
var wordsCount = wordsInDoc.Count(w => w.Equals(allWords[j]));
docsWordsArray[i, j] = wordsCount;
}
}
var docsWordsMatrix = Matrix<double>.Build.DenseOfArray(docsWordsArray);
wordsCorrelationMatrix = docsWordsMatrix.Transpose().Multiply(docsWordsMatrix).NormalizeRows(1.0);
#endregion
//NextWordCounting
#region Next Word Counting
for (int i = 0; i < _rawDocuments.Count; i++)
{
var words = GetWords(_rawDocuments[i]);
for (int j = 0; j < words.Length; j++)
{
if (!_nextWordCounter.ContainsKey(words[j]))
{// jeszcze nie było takiego słowa
_nextWordCounter[words[j]] = new Dictionary<string, int>();
if (j + 1 < words.Length)
{//istnieje następne słowo
_nextWordCounter[words[j]][words[j + 1]] = 1;
}
}
else
{// było już i incrementujemy wartość jego następnego słowa
if (j + 1 < words.Length)
{// istnieje następne słowo
if (!_nextWordCounter[words[j]].ContainsKey(words[j + 1]))
{//następnego słowa jeszcze nie ma w słowniku
_nextWordCounter[words[j]][words[j + 1]] = 1;
}
else
{//następne słowo już jest w słowniku - incrementujemy wartość
_nextWordCounter[words[j]][words[j + 1]]++;
}
}
}
}
}
#endregion
//LSI Calculating
#region LSI
double[,] termsDocsArray = new double[_rawKeywords.Count, _documents.Count];
//bag of words
for (int i = 0; i < _rawKeywords.Count; i++)
{
for (int j = 0; j < _documents.Count; j++)
{
termsDocsArray[i, j] = _documents[j].BagOfWords[_rawKeywords[i]];
}
}
// TF/IDF
//for (int i = 0; i < _rawKeywords.Count; i++)
//{
// for (int j = 0; j < _documents.Count; j++)
// {
// termsDocsArray[i, j] = _documents[j].TF[_rawKeywords[i]] * IDF[_rawKeywords[i]];
// }
//}
Matrix<double> termsDocsMatrix = Matrix<double>.Build.DenseOfArray(termsDocsArray);
var svd = termsDocsMatrix.Svd();
var K = svd.U;
var S = Matrix<double>.Build.DenseOfDiagonalArray(svd.S.ToArray());
var DT = svd.VT.SubMatrix(0, S.RowCount, 0, svd.VT.ColumnCount);
var D = DT.Transpose();
S_s = S.SubMatrix(0, S.RowCount - reductionCount, 0, S.ColumnCount - reductionCount);
K_s = K.SubMatrix(0, K.RowCount, 0, K.ColumnCount - Math.Abs(K.ColumnCount - S_s.ColumnCount));
D_sT = DT.SubMatrix(0, DT.RowCount - reductionCount, 0, DT.ColumnCount);
D_s = D_sT.Transpose();
LsiMatrix = K_s * S_s * D_sT;
#endregion
}
