public void Start()
{
if (!outputpath.EndsWith("\\"))
{
outputpath += "\\";
}
var tokenizerConfig = new TokenizeConfig(tokenizeConfigStr);
var searcher = LuceneOperations.GetIndexSearcher(inputpath);
var max_doc_num = (int)(searchDocRatio * searcher.GetIndexReader().NumDocs());
var scoredDocs = LuceneOperations.Search(searcher, searchfield, keywords, max_doc_num);
int iter = 0;
bool bContinue = threshold == 0 ? false : true;
while (bContinue && iter < 5)
{
iter++;
Console.WriteLine("iteration------------------" + iter);
List <string> keywordsNew;
#region Calculate Keywords
var counter = new Counter <string>();
foreach (var scoredDoc in scoredDocs)
{
var doc = searcher.Doc(scoredDoc.doc);
var content = doc.Get(searchfield);
foreach (var word in NLPOperations.Tokenize(content, tokenizerConfig))
{
counter.Add(word);
}
}
keywordsNew = counter.GetMostFreqObjs(keywordNum);
#endregion
var scoredDocsNew = LuceneOperations.Search(searcher, searchfield, keywordsNew, max_doc_num);
#region Test whether exit
int repeatNum = 0;
var docIDs = new HashSet <int>();
foreach (var scoredDoc in scoredDocs)
{
docIDs.Add(scoredDoc.doc);
}
foreach (var scoredDocNew in scoredDocsNew)
{
if (docIDs.Contains(scoredDocNew.doc))
{
repeatNum++;
}
}
bContinue = (double)repeatNum / scoredDocs.Length < threshold;
#endregion
Console.WriteLine(repeatNum + " " + scoredDocsNew.Length);
keywords = keywordsNew;
scoredDocs = scoredDocsNew;
Console.WriteLine(StringOperations.GetMergedString(keywords));
}
max_doc_num = (int)(saveDocRatio * searcher.GetIndexReader().NumDocs());
scoredDocs = LuceneOperations.Search(searcher, searchfield, keywords, max_doc_num);
var writer = LuceneOperations.GetIndexWriter(outputpath);
foreach (var scoredDoc in scoredDocs)
{
Document doc = searcher.Doc(scoredDoc.doc);
writer.AddDocument(doc);
}
writer.Optimize();
writer.Close();
if (isPrintRemovedDocuments)
{
var sw = new StreamWriter(outputpath + "removeDocuments.txt");
var selectedDocIDs = new HashSet <int>();
foreach (var scoredDoc in scoredDocs)
{
selectedDocIDs.Add(scoredDoc.doc);
}
var reader = searcher.GetIndexReader();
for (int iDoc = 0; iDoc < reader.NumDocs(); iDoc++)
{
if (!selectedDocIDs.Contains(iDoc))
{
sw.WriteLine(LuceneOperations.GetDocumentString(reader.Document(iDoc)));
}
}
reader.Close();
sw.Flush();
sw.Close();
}
searcher.Close();
Console.WriteLine("Done");
Console.ReadKey();
}
private string DocumentToString(Document document)
{
return(LuceneOperations.GetDocumentString(document));
}
