public static void ClearAll()
{
using (var _db = new EngineDBDataContext())
{
_db.Terms.DeleteAllOnSubmit(_db.Terms);
_db.Scores.DeleteAllOnSubmit(_db.Scores);
_db.SubmitChanges();
}
}
public static void RemoveStory(int storyID)
{
using (var _db = new EngineDBDataContext())
{
_db.Terms.DeleteAllOnSubmit(_db.Terms.Where(t => t.StoryID == storyID));
_db.Scores.DeleteAllOnSubmit(_db.Scores.Where(s => s.Story1ID == storyID || s.Story2ID == storyID));
_db.SubmitChanges();
}
}
/// <summary>
/// Store a new document or update an existing one
/// </summary>
/// <param name="documentID"></param>
/// <param name="body"></param>
public static void StoreStory(int storyID, string text)
{
using (var _db = new EngineDBDataContext())
{
IList <Term> terms = TermSet(text);
// grab idfs
var idfs = from term in _db.Terms
where term.StoryID != storyID
group term by term.Text
into g
where terms.Select(t => t.Text).Contains(g.Key)
select new { TermVal = g.Key, Count = g.Count() };
// calculate tfidfs
double num_docs = (from term in _db.Terms
where term.StoryID != storyID
group term by term.StoryID
into g
select g).Count() + 1;
foreach (Term term in terms)
{
term.StoryID = storyID;
Term term1 = term;
var idf = 1;
// need to refactor this because it's doing a massive SELECT * FROM tbl WHERE x IN ...
var idfEntry = idfs.SingleOrDefault(t => t.TermVal.ToLower() == term1.Text.ToLower());
if (idfEntry != null)
{
idf = idfEntry.Count;
}
term.tfidf = term.tf * Math.Log(num_docs / idf);
}
// delete old entries
_db.Terms.DeleteAllOnSubmit(_db.Terms.Where(t => t.StoryID == storyID));
// store
_db.Terms.InsertAllOnSubmit(terms);
_db.SubmitChanges();
}
ScoreStory(storyID);
}
/* Calculate the cosine similarity measure of ID versus all other documents in the corpus */
private static void ScoreStory(int storyID)
{
// consider passing data context as an argument to the function
using (var _db = new EngineDBDataContext())
{
// delete all existing scores for documentID
_db.Scores.DeleteAllOnSubmit(from score in _db.Scores
where score.Story1ID == storyID || score.Story2ID == storyID
select score);
_db.SubmitChanges();
// grab the principle document's terms/tfidfs
var terms = from term in _db.Terms
where term.StoryID == storyID
select new { Text = term.Text, tfidf = term.tfidf };
// magnitude of the document's vector
var mag1 = Math.Sqrt(terms.Sum(t => t.tfidf));
// ids of documents to score against
var ids = from term in _db.Terms
where term.StoryID != storyID
group term by term.StoryID
into g
select g.Key;
// score against each document
foreach (int id in ids)
{
var id1 = id;
// grab the other document's terms/tfidfs
var terms2 = from term in _db.Terms
where term.StoryID == id1
select new { Text = term.Text, tfidf = term.tfidf };
var mag2 = Math.Sqrt(terms2.Sum(t => t.tfidf));
// calculate the dot product
var dot_product = 0.0;
foreach (var term in terms)
{
var term1 = term;
var term2 = terms2.SingleOrDefault(t => t.Text.ToLower() == term1.Text.ToLower());
if (term2 != null)
{
dot_product += term1.tfidf * term2.tfidf;
}
}
var euclidean_dist = mag1 * mag2;
// linq may complain about overwriting existing keys... we'll see. if so, delete scores before InsertOnSubmit()
var score = new Score
{
Story1ID = storyID,
Story2ID = id,
Value = dot_product / euclidean_dist
};
if (double.IsNaN(score.Value))
{
score.Value = 0;
}
_db.Scores.InsertOnSubmit(score);
}
_db.SubmitChanges();
//
}
}