public TermExtractor()
{
termCollectors = new Hashtable();
stopList = new d.StopList();
stemmer = new PorterStemmer();
parseHelper = new ParseHelper();
delims = parseHelper.GetDelims();
}
public TermParser()
{
stopList = new d.StopList();
delims = parseHelper.GetDelims();
stemmer = new PorterStemmer();
termData = new d.TermData();
DocData = new d.DocData();
termdocData = new d.TermDocData();
parseHelper = new ParseHelper();
}
private void loadText(string text)
{
d.StopList stoplist = new d.StopList();
PorterStemmer stemmer = new PorterStemmer();
ParseHelper ph = new ParseHelper();
text = text.ToLower();
char[] delims = ph.GetDelims();
string[] temp = text.Split(delims);
Hashtable terms = new Hashtable();
string term;
for (int i = 0; i < temp.Length; i++)
{
term = stemmer.stemTerm(temp[i]);
if (!stoplist.Contains(term))
{
if (terms.Contains(term))
{
terms[term] = Convert.ToInt32(terms[term]) + 1;
}
else
{
terms.Add(term, 1);
}
}
}
StringBuilder sbt = new StringBuilder();
StringBuilder sbw = new StringBuilder();
IDictionaryEnumerator en = terms.GetEnumerator();
string t;
string w;
while (en.MoveNext())
{
t = en.Key.ToString();
w = terms[t].ToString();
sbt.AppendFormat("{0},", t);
sbw.AppendFormat("{0},", w);
}
queryTerms = sbt.ToString();
queryWeights = sbw.ToString();
}
public void AddAnchorText()
{
d.StopList stopList = new d.StopList();
ParseHelper parseHelper = new ParseHelper();
char[] delims = parseHelper.GetDelims();
PorterStemmer stemmer = new PorterStemmer();
d.LinkData ld = new d.LinkData();
d.TermDocData tdd = new d.TermDocData();
DataTable linksTable;
int docId;
StringBuilder sb;
string[] terms;
string term;
Hashtable currTerms;
DataTable dt = new d.DocData().GetIds();
for (int i = 0; i < dt.Rows.Count; i++)
{
if (i % 10 == 0)
{
Console.WriteLine(i);
}
//accumulate all link text for this doc into StringBuilder
sb = new StringBuilder();
docId = (int)dt.Rows[i][0];
linksTable = ld.GetRecordsByToId(docId);
foreach (DataRow dr in linksTable.Rows)
{
sb.AppendFormat("{0} ", dr[0].ToString());
}
//accum terms + counts into currTerms hashtable
currTerms = new Hashtable();
terms = sb.ToString().Split(delims);
for (int j = 0; j < terms.Length; j++)
{
term = stemmer.stemTerm(terms[j].ToLower().Trim());
if (term != "home" && term.Length > 0 && term.Length < 25 && !stopList.Contains(term) && parseHelper.IsAsciiLetters(term))
{
if (!currTerms.Contains(term))
{
currTerms.Add(term, 1);
}
else
{
currTerms[term] = (int)currTerms[term] + 1;
}
}
}
//write terms and counts to database
IDictionaryEnumerator en = currTerms.GetEnumerator();
string currTerm;
int currCount;
while (en.MoveNext())
{
currTerm = en.Key.ToString();
currCount = (int)currTerms[currTerm];
tdd.UpdateAnchorTextCount(currTerm, docId, currCount);
}
}
}