private SearchEngine ChooseApproriateSearchEngine(string query)
{
if (query.Equals(Unicode2ASCII.Convert(query)))
{
return(this.mySearchEngine2);
}
return(this.mySearchEngine);
}
// Perform indexing using SPIMI
// All files generated during the indexing process are stored in SPIMI folder
// Return value: the inverted index of the collection
public int Index()
{
Console.WriteLine("Index Construction using Single-pass in-memory Indexing");
// Initialize the dictionary
Dictionary <string, List <int> > dict = new Dictionary <string, List <int> >(252143);
// Count the size (in bytes) of the values (term, docId, frequency) added to dictionary and postings lists
int count = 0;
// Count the number of blocks written to disk
int blockId = 0;
// Delete old files
string[] filePaths = Directory.GetFiles("..//..//SPIMI//");
foreach (string filePath in filePaths)
{
if (filePath.Contains("index.txt") || filePath.Contains("index_map.txt") || filePath.Contains("index_length.txt"))
{
continue;
}
File.Delete(filePath);
}
// Count the total number of documents
int docId = 0;
// Read and process all documents
foreach (string docPath in Directory.GetFiles(path))
{
using (StreamReader sr = new StreamReader(docPath))
{
string content = sr.ReadToEnd().ToLower();
if (content != null)
{
content = Unicode2ASCII.Convert(content);
SPIMI_Invert(content, docId, ref dict, ref count, "SPIMIvi_", ref blockId);
}
++docId;
}
}
// Write the remaining postings lists
if (dict.Count > 0)
{
Console.WriteLine("Block " + blockId.ToString() + ": Finished!");
List <string> sortedTerm = dict.Keys.ToList();
sortedTerm.Sort();
WriteBlockToDisk(sortedTerm, dict, "SPIMIvi_" + blockId.ToString());
++blockId;
}
// Merge all blocks
Console.WriteLine("Merging blocks...");
MergeBlocks("SPIMIvi_", blockId, "..//..//SPIMI//indexvi", docId);
Console.WriteLine("Inverted index successfully constructed.");
return(docId);
}
protected override string GetDocContent(int docId)
{
string content = "";
string[] fileNames = Directory.GetFiles(docPath);
using (StreamReader rd = new StreamReader(fileNames[docId]))
{
content = rd.ReadToEnd();
}
content = Unicode2ASCII.Convert(content);
return(content);
}
// Read all stopwords from file
private HashSet <string> GetStopwords()
{
IEnumerable <string> stopwords = File.ReadLines(path + "//Stopword//stopwords_vi.txt");
string[] stopwordarray = stopwords.ToArray <string>();
for (int i = 0; i < stopwordarray.Length; ++i)
{
stopwordarray[i] = Unicode2ASCII.Convert(stopwordarray[i]);
}
return(new HashSet <string>(stopwordarray));
}
protected override Dictionary <string, int> Preprocess(string query)
{
query = Unicode2ASCII.Convert(query);
// Tokenize the query
MatchCollection words = Tokenizer.TokenizeDoc(query, @"[a-zA-Z]+");
// Get all terms and their frequencies in the query
Dictionary <string, int> terms = new Dictionary <string, int>();
foreach (var word in words)
{
string term = word.ToString().ToLower();
if (terms.ContainsKey(term))
{
++terms[term];
}
else
{
terms.Add(term, 1);
}
}
return(terms);
}
