public void Process(string categoryFile, string postagFile,string gramCategoryFile)
{
int order = 0;
Dictionary<int, string> orderCategory = new Dictionary<int, string>();
string currentLine;
StreamReader reader = new StreamReader(categoryFile);
while ((currentLine = reader.ReadLine()) != null)
{
string category = currentLine.Split(' ')[0];
if (category != "NUM:date")
category = category.Split(':')[0];
orderCategory.Add(order, category);
order++;
}
reader.Close();
reader = new StreamReader(postagFile);
CFormalizeString stringFormalizor = new CFormalizeString();
StreamWriter writer = new StreamWriter(gramCategoryFile);
Dictionary<string, Dictionary<string, int>> bigramCategoryF = new Dictionary<string, Dictionary<string, int>>();
order=0;
while ((currentLine = reader.ReadLine()) != null)
{
string fullLine = currentLine;
while (fullLine.Contains('?') == false)
{
currentLine = reader.ReadLine();
fullLine = fullLine + " " + currentLine;
}
string[] tokens = fullLine.ToLower().Split(' ');
string bigram="";
string category=orderCategory[order];
order++;
if (tokens[0].StartsWith("what"))
{
foreach(string token in tokens)
if (token.Contains("_n"))
{
bigram="what"+" "+token.Split('_')[0];
break;
}
}
else
{
bigram = tokens[0].Split('_')[0] + " " + tokens[1].Split('_')[0];
}
if (bigram!="")
{
bigram = stringFormalizor.FormalizeString(bigram);
if (bigramCategoryF.ContainsKey(bigram) == false)
bigramCategoryF.Add(bigram,new Dictionary<string,int>());
if (bigramCategoryF[bigram].ContainsKey(category) == false)
bigramCategoryF[bigram].Add(category, 0);
bigramCategoryF[bigram][category]++;
}
}
reader.Close();
foreach (string bigram in bigramCategoryF.Keys)
{
int maxF = 0;
string maxCategory = "";
foreach (string category in bigramCategoryF[bigram].Keys)
if (bigramCategoryF[bigram][category] > maxF)
{
maxF = bigramCategoryF[bigram][category];
maxCategory = category;
}
writer.WriteLine(bigram + "\t" + maxCategory);
}
writer.Close();
}
void GenerateFamousIdSet(string inputFile)
{
StreamReader reader = new StreamReader(inputFile);
string currentLine;
CFormalizeString sf=new CFormalizeString();
int lineCount = 0;
while ((currentLine = reader.ReadLine()) != null)
{
lineCount++;
if (lineCount % 1000000 == 0)
{
Console.WriteLine("depth:{0}\tline:{1}\tcount:{2}", depth, lineCount,idPathList.Count);
//break;
}
string[] tokens=currentLine.Split('\t');
if ((tokens[1]=="<type.object.name>" || tokens[1] == "<common.topic.alias>" )
&& famousEntitySet.Contains(tokens[2].ToLower())
&& IsId(tokens[0]))
{
if (idPathList.ContainsKey(tokens[0])==false)
idPathList.Add(tokens[0],new List<string>());
idPathList[tokens[0]].Add(tokens[2].ToLower() + "\t");
}
}
reader.Close();
}