public virtual Document PrepareDocument(string contents, IReadOnlyCollection <string> keywords)
{
var singleLine = contents.Replace("\r", "").Replace("\n", " ");
var cleaned = RemoveSpecialCharacters(singleLine);
var lower = cleaned.ToLower();
var splitted = lower.Split();
var stemmer = new PorterStemmer();
var stemmed = splitted.Select(stemmer.StemWord).ToArray();
var vector = this.DocumentToVector(stemmed, keywords);
return(new Document(contents, stemmed, vector));
}
public SearchEngine(string[] keywords, IDocumentSanitizer sanitizer)
{
this.sanitizer = sanitizer;
this.stemmer = new PorterStemmer();
this.documents = new List <Document>();
this.keywords = new HashSet <string>();
this.inverseDocumentFrequencies = new List <double>();
this.feedbackCalculator = new RocchioFeedbackCalculator();
foreach (var keyword in keywords)
{
var stemmed = this.stemmer.StemWord(keyword);
this.keywords.Add(stemmed);
}
this.clusteringStrategy = new KMeansClustering(9, 100, new Random());
}