public Hashtable preprocessTranslationModel(string[] originalFiles, string[] summariesFiles)
{
DocsStatistics originalDocStats = processFiles(originalFiles);
DocsStatistics summariesDocStats = processFiles(summariesFiles);
Hashtable translationModel = null;
foreach (string word in summariesDocStats.wordsCount.Keys)
{
if (originalDocStats.wordsCount[word] != null)
{
continue;
}
double originalCount = (double)((originalDocStats.wordsCount[word] == null) ? 0 : originalDocStats.wordsCount[word]);
double summaryCount = (double)((summariesDocStats.wordsCount[word] == null) ? 0 : summariesDocStats.wordsCount[word]);
if (translationModel == null)
{
translationModel = new Hashtable();
}
translationModel[word] = summaryCount / originalCount;
}
return(translationModel);
}
private void TrainingFilesDialog_FileOk(object sender, CancelEventArgs e)
{
string[] fileNames = this.TrainingFilesDialog.FileNames;
ArrayList docs = new ArrayList();
this.progressBar.Show();
this.progressBar.Minimum = 0;
this.progressBar.Maximum = fileNames.Length + (fileNames.Length / 4);
this.progressBar.Value = 0;
foreach (string fileName in fileNames)
{
string fileText = File.ReadAllText(fileName, Encoding.Default);
Document doc = docProcessor.process(fileText);
docs.Add(doc);
this.progressBar.Increment(1);
}
this.trainingDocs = docs;
this.docsStat = DocsStatistics.generateStatistics(docs);
this.progressBar.Value = this.progressBar.Maximum;
this.progressBar.Hide();
this.AlgorithmCmbo.Enabled = true;
}
private void load(string clusterDir)
{
DocumentProcessor docProcessor = new DocumentProcessor();
ArrayList docs = new ArrayList();
string[] clusterFiles = Directory.GetFiles(clusterDir, "*.txt", SearchOption.TopDirectoryOnly);
foreach (string filename in clusterFiles)
{
string fileText = File.ReadAllText(filename, Encoding.Default);
Document doc = docProcessor.process(fileText);
docs.Add(doc);
}
DocsStatistics docStats = DocsStatistics.generateStatistics(docs);
Hashtable centroid = new Hashtable();
foreach (string word in docStats.wordsCount.Keys)
{
//centroid[firstWord] = (((int)docStats.wordsCount[firstWord]) * idf(docStats, firstWord)) / docs.Count;
centroid[word] = (((int)docStats.wordsCount[word]) * IDF.getInstance().get(word)) / docs.Count;
}
this.centroidWords = applyKeepWords(centroid, this.keepWords);
}
private static void processFile(DocsStatistics docStats, string filename)
{
DocumentProcessor docProcessor = new DocumentProcessor();
string fileContent = File.ReadAllText(filename, Encoding.Default);
using (Document doc = docProcessor.process(fileContent))
{
docStats.addDocument(doc);
}
}
public static double termFrequency(DocsStatistics docStats, string word)
{
//double tf = sent.wordsCount[firstWord] == null ? 0 : ((int)sent.wordsCount[firstWord] / sent.words.Length);
double tf = docStats.wordsCount[word] == null ? 0 : (int)docStats.wordsCount[word];
if (tf != 0)
{
tf = tf / ((HashSet <Document>)docStats.wordRefs[word]).Count;
}
return(tf);
}
private DocsStatistics processFiles(string[] files)
{
DocumentProcessor docProcessor = new DocumentProcessor();
DocsStatistics docStats = new DocsStatistics();
foreach (string filename in files)
{
Document doc = docProcessor.process(filename);
docStats.addDocument(doc);
}
return(docStats);
}
private void load(ArrayList docs)
{
DocsStatistics docStats = DocsStatistics.generateStatistics(docs);
this.idf = new Hashtable();
foreach (string word in docStats.wordsCount.Keys)
{
double wordRefCount = docStats.wordRefs[word] == null ? 0 : ((HashSet <Document>)docStats.wordRefs[word]).Count;
double wordIdf = Math.Log(docStats.docCount / (wordRefCount + 1));
this.idf[word] = wordIdf;
}
}
public static IDF fromFiles(string[] files)
{
DocsStatistics docStats = new DocsStatistics();
DocumentProcessor docProcessor = new DocumentProcessor();
int i = 0;
foreach (string file in files)
{
++i;
//processFile(docStats, file);
//*
string fileContent = File.ReadAllText(file, Encoding.Default);
Document doc = docProcessor.process(fileContent);
docStats.addDocument(doc);
/*
* if ((i % 1000) == 0)
* {
* System.GC.Collect();
* Trace.write("Done for : " + i);
* }
* //*/
//*/
//doc = null;
}
IDF idf = new IDF();
foreach (string word in docStats.wordsCount.Keys)
{
//double wordRefCount = docStats.wordRefs[firstWord] == null ? 0 : ((HashSet<Document>)docStats.wordRefs[firstWord]).Count;
double wordRefCount = docStats.wordRefsCount[word] == null ? 0 : ((int)docStats.wordRefsCount[word]);
double wordIdf = Math.Log(docStats.docCount / (wordRefCount));
idf.idf[word] = wordIdf;
}
return(idf);
}
public ArrayList buildCentroids(ArrayList docs, IDF idfdb)
{
ArrayList centroids = new ArrayList();
foreach (Document doc in docs)
{
ArrayList currDoc = new ArrayList();
currDoc.Add(doc);
DocsStatistics currDocStats = DocsStatistics.generateStatistics(currDoc);
Hashtable docVector = new Hashtable();
foreach (DictionaryEntry entry in currDocStats.wordsCount)
{
string word = (string)entry.Key;
int count = (int)entry.Value;
//double idf = CentroidAlgorithm2.idf(allDocStats, firstWord);
double idf = idfdb.get(word);
if (idf < this.idfThreshold)
{
continue;
}
double tfidf = ((double)count) * idf;
docVector[word] = tfidf;
}
if (centroids.Count == 0)
{
Centroid centroid = new Centroid(docVector, this.keepWords);
centroid.noOfDocuments = 1;
centroids.Add(centroid);
}
else
{
Centroid nearestCentroid = null;
double maxSimilarity = double.MinValue;
foreach (Centroid centroid in centroids)
{
double similarity = sim(IDF.getInstance(), centroid.values, docVector);
if (similarity > simThreshold)
{
if (similarity > maxSimilarity)
{
maxSimilarity = similarity;
nearestCentroid = centroid;
}
}
}
if (nearestCentroid == null)
{
nearestCentroid = new Centroid(docVector, this.keepWords);
centroids.Add(nearestCentroid);
}
else
{
nearestCentroid.addDocument(docVector);
}
}
}
// Apply the KEEP_WORDS parameter for each centroid
/*
* foreach (Centroid centroid in centroids)
* {
* Hashtable centroidValues = centroid.values;
*
* DictionaryEntry[] centValuesArr = new DictionaryEntry[centroids.Count];
*
* centroidValues.CopyTo(centValuesArr, 0);
*
* Array.Sort(centValuesArr, new DictionaryEntryValueComparer());
* Array.Reverse(centValuesArr);
*
* DictionaryEntry[] finalCentroidValuesArr = new DictionaryEntry[this.keepWords];
*
* Array.Copy(centValuesArr, finalCentroidValuesArr, this.keepWords);
*
* Hashtable finalCentroidValues = new Hashtable();
*
* foreach (DictionaryEntry entry in finalCentroidValuesArr)
* {
* finalCentroidValues.Add(entry.Key, entry.Value);
* }
*
* centroid.values = finalCentroidValues;
* }
* //*/
//*
foreach (Centroid centroid in centroids)
{
centroid.applyKeepWords();
}
//*/
// Trace
/*
* int i = 0;
* foreach (Centroid centroid in centroids)
* {
* Trace.write("Centroid #" + (++i));
* foreach (DictionaryEntry entry in centroid.values)
* {
* Trace.write(entry.Key + " : " + entry.Value);
* }
* }
* //*/
return(centroids);
}
public static double idf(DocsStatistics docStats, string word)
{
double wordRefCount = docStats.wordRefs[word] == null ? 0 : ((HashSet <Document>)docStats.wordRefs[word]).Count;
return(Math.Log(docStats.docCount / (wordRefCount + 1)));
}
