private void LoadUnigrams(UniGramFile uniGramFiles, List <List <string> > ngramLists)
{
Ngram ngram;
string lineWordsFormated;
while ((ngram = uniGramFiles.Next()) != null)
{
lineWordsFormated = string.Join(" ", ngram.Words);
foreach (string w in ngram.Words)
{
string nonDiacriticsWord = StringRoutines.MyDiacriticsRemover(w);
List <string> foundList = trie.Find(nonDiacriticsWord);
if (foundList == null)
{
var l = new List <string> {
lineWordsFormated
};
ngramLists.Add(l);
trie.Add(nonDiacriticsWord, l);
}
else
{
foundList.Add(lineWordsFormated);
}
}
}
}
private void InsertWordsSqlBulkCopy(UniGramFile file, DiacriticsDBEntities db)
{
var dtWords = new DataTable();
dtWords.Columns.Add("Id");
dtWords.Columns.Add("Value");
file.ReOpen();
Ngram ng;
int id = 0;
var counter = 0;
while ((ng = file.Next()) != null)
{
string nonDiacriticsW = StringRoutines.MyDiacriticsRemover(ng.ToString());
dtWords.Rows.Add(++id, nonDiacriticsW);
if (++counter % 100000 == 0)
{
Console.WriteLine(counter + " words prepared for insertion.");
}
}
InsertIntoDb(dtWords, db, "dbo.Words");
}
private void InsertUnigramsSqlBulkCopy(UniGramFile file, DiacriticsDBEntities db)
{
var dtUniGrams = new DataTable();
dtUniGrams.Columns.Add("Word1");
dtUniGrams.Columns.Add("WordId");
dtUniGrams.Columns.Add("Id");
dtUniGrams.Columns.Add("Frequency");
file.ReOpen();
Ngram ng;
int wordId;
var uniGramId = 0;
var counter = 0;
while ((ng = file.Next()) != null)
{
string w = ng.ToString();
string nonDiacriticsW = StringRoutines.MyDiacriticsRemover(w);
wordId = wordTrie.Find(nonDiacriticsW);
if (wordId != 0)
{
dtUniGrams.Rows.Add(w, wordId, ++uniGramId, ng.Frequency);
}
else
{
throw new Exception("Word '" + nonDiacriticsW + "' is not present in Trie!");
}
if (++counter % 10000 == 0)
{
Console.WriteLine(counter + " unigrams prepared for insertion.");
}
}
InsertIntoDb(dtUniGrams, db, "dbo.UniGramEntities");
}
