/// <summary>
/// Insert new documetn in storage.
/// </summary>
/// <param name="documentTermsData"></param>
/// <returns>Return new Id of new document </returns>
public string PostDocumentTerms(DocumentTermsData documentTermsData)
{
lock (_lockerDocumentTermsColl)
{
return(DocumentTermsColl.Insert(documentTermsData));
}
}
/// <summary>
/// Reduce number for each count in TermDocumentCountColl {term, count}
/// Delete {Document, List<TermData>} from DocumentTermsColl
/// ?Delete { term, documentId} from TermDocumentColl
/// 2020-12-23T10:12:38
/// </summary>
/// <param name="documentId"></param>
/// <returns></returns>
public bool DeleteDocument(string documentId)
{
lock (_lockerDocumentTermsColl)
{
DocumentTermsData documentTermsData = DocumentTermsColl.FindOne(d => d.Document == documentId);
if (documentTermsData != null)
{
//Reduce number for each count in TermDocumentCountColl {term, count}
lock (_lockerTermDocumentCountColl)
{
foreach (TermData termData in documentTermsData.Terms)
{
TermDocumentCountData termDocumentCountData = TermDocumentCountColl.FindOne(d => d.Term == termData.Term);
if (termDocumentCountData != null)
{
termDocumentCountData.Count--;
TermDocumentCountColl.Update(termDocumentCountData);
}
}
//Delete {Document, List<TermData>} from DocumentTermsColl
return(DocumentTermsColl.Delete(documentTermsData.Id));
}
}
return(false);
}
}
/// <summary>
/// Adds a new document along with its terms to the database.
/// </summary>
/// <param name="document"></param>
/// <param name="documentTerms"></param>
public void AddDocument(string documentId, List <TermData> documentTerms)
{
DocumentTermsData documentTermsData = new DocumentTermsData()
{
Document = documentId,
Terms = documentTerms
};
Storage.PostDocumentTerms(documentTermsData);
Storage.PutTermDocumentCounts(documentTerms);
}
/// <summary>
/// Gets the cosine similarity of vectors of keywords of two documents.
/// Tf-Idf and Cosine similarity: https://janav.wordpress.com/2013/10/27/tf-idf-and-cosine-similarity/
/// </summary>
/// <param name="documentId1"></param>
/// <param name="documentId2"></param>
/// <param name="tfIdfEstimator"></param>
/// <returns>Cosine similarity</returns>
public static double GetDocumentSimilarityExt(DocumentTermsData documentTermsData1, string documentId2, TfIdfEstimatorExt tfIdfEstimator)
{
//using var db = new LiteDatabase(tfIdfEstimator.TfIdfStorage.ConnectionString);
//var documentTermsColl = db.GetCollection<DocumentTermsData>(tfIdfEstimator.TfIdfStorage.DocumentTermsColl);
// Get all keywords from the two documents and make a union of them.
DocumentTermsData doc1 = documentTermsData1;//2020-12-28T08:49:23 tfIdfEstimator.Storage.DocumentTermsColl.FindOne(x => x.Document == documentId1);
DocumentTermsData documentTermsData2 = tfIdfEstimator.Storage.DocumentTermsColl.FindOne(x => x.Document == documentId2);
List <TermData> terms1 = doc1.Terms;
List <TermData> terms2 = documentTermsData2.Terms;
var set = new HashSet <string>();
foreach (var keyword in terms1)
{
set.Add(keyword.Term);
}
foreach (var keyword in terms2)
{
set.Add(keyword.Term);
}
// Get term scores of keywords in a union for both documents.
List <double> list1 = new List <double>();
List <double> list2 = new List <double>();
foreach (var keyword in set)
{
list1.Add(tfIdfEstimator.GetOneTermInDocument(doc1.Document, keyword).TermScore);
list2.Add(tfIdfEstimator.GetOneTermInDocument(documentId2, keyword).TermScore);
}
// Calculate the cosine similarity. CosSimilarity(v1, v2) = dot(v1,v2) / (norm(v1) * norm(v2)) where v1 and v2 are vectors
double dot = DotProduct(list1, list2);
double norm1 = 0;
double norm2 = 0;
foreach (var item in list1)
{
norm1 += Math.Pow(item, 2);
}
foreach (var item in list2)
{
norm2 += Math.Pow(item, 2);
}
norm1 = Math.Sqrt(norm1);
norm2 = Math.Sqrt(norm2);
double similarity = (norm1 * norm2 == 0 ? 0 : dot / (norm1 * norm2));
return(similarity);
}
/// <summary>
/// Returns top N most similar documents to the first document along with its cosine similarities.
/// </summary>
/// <param name="documentId1"></param>
/// <param name="numberOfDocuments"></param>
/// <param name="tfIdfEstimator"></param>
/// <returns>List of most similar documents</returns>
public static List <DocumentSimilarityScoreData> GetSimilarDocumentsOLD1(string documentId1, int numberOfDocuments, TfIdfEstimatorExt tfIdfEstimator)
{
IEnumerable <DocumentTermsData> docs = tfIdfEstimator.Storage.DocumentTermsColl.FindAll();//.ToList();//TODO: optimize it!!! - 2020-12-22T10:04:31
DocumentTermsData documentTermsData1 = tfIdfEstimator.Storage.DocumentTermsColl.FindOne(x => x.Document == documentId1);
// Get similarity scores of all those documents.
var docScores = new List <DocumentSimilarityScoreData>();
foreach (DocumentTermsData documentTermsData in docs)
{
if (documentTermsData1.Document == documentTermsData.Document)
{
continue;//skip it
}
docScores.Add(new DocumentSimilarityScoreData()
{
Document = documentTermsData.Document,
Score = GetDocumentSimilarityExt(documentTermsData1, documentTermsData.Document, tfIdfEstimator)
});
}
//Order them by their scores descending and return top N of them.
docScores.OrderByDescending(x => x.Score);
return(docScores.GetRange(0, numberOfDocuments));
/*OLD CODE: 2020-12-28T08:57:09
* // Get all docs except the one that we are looking the most similar to.
* var docs = tfIdfEstimator.Storage.DocumentTermsColl.FindAll().ToList();//TODO: optimize it!!! - 2020-12-22T10:04:31
* docs.Remove(docs.Find(x => x.Document == document1));
*
* // Get similarity scores of all those documents.
* var docScores = new List<DocumentSimilarityScoreData>();
* foreach (var doc in docs)
* {
* docScores.Add(new DocumentSimilarityScoreData()
* {
* Document = doc.Document,
* Score = GetDocumentSimilarityExt(document1, doc.Document, tfIdfEstimator)
* });
* }
*
* //Order them by their scores descending and return top N of them.
* docScores.OrderByDescending(x => x.Score);
* return docScores.GetRange(0, numberOfDocuments);
*/
}
/// <summary>
/// Takes a names of the document from which to get a specific term with its tf-idf value.
/// </summary>
/// <param name="documentId"></param>
/// <param name="term"></param>
/// <returns>Returns a TermsScoreData object</returns>
public TermScoreData GetOneTermInDocument(string documentId, string term)
{
DocumentTermsData documentTermsData = Storage.DocumentTermsColl.FindOne(x => x.Document == documentId);
long countOfTerms = documentTermsData.Terms.Sum(x => x.Count);
var term2 = documentTermsData.Terms.Find(x => x.Term == term);
long countOfTerm = (term2 == null ? 0 : term2.Count);
double termFrequency = (countOfTerms == 0 ? 0 : countOfTerm / (double)countOfTerms);
int countOfDocs = Storage.DocumentTermsColl.Count();
long countOfDocsWithTerm = Storage.TermDocumentCountColl.FindOne(x => x.Term == term).Count;
double inverseDocumentFrequencySmooth = Math.Log10(countOfDocs / (double)(countOfDocsWithTerm + 1d)) + 1d;
double tfidfValue = termFrequency * inverseDocumentFrequencySmooth;
var tsd = new TermScoreData
{
Term = term,
TermScore = tfidfValue
};
return(tsd);
}
/// <summary>
/// Returns top N most similar documents to the first document along with its cosine similarities.
/// </summary>
/// <param name="documentId1"></param>
/// <param name="numberOfDocuments"></param>
/// <param name="tfIdfEstimator"></param>
/// <returns>List of most similar documents</returns>
public static List <DocumentSimilarityScoreData> GetSimilarDocuments(string documentId1, int numberOfDocuments, TfIdfEstimatorExt tfIdfEstimator)
{
DocumentTermsData documentTermsData1 = tfIdfEstimator.Storage.DocumentTermsColl.FindOne(x => x.Document == documentId1);
//bool Compare(string term)//local function
//{
// TermData termData = documentTermsData1.Terms.Find(t => t.Term == term);
// if (termData != null)
// {
// return true;
// }
// return false;
//}
//Func<string, bool> Predicate = Compare;
//Example using multi key index https://github.com/mbdavid/LiteDB/blob/master/LiteDB.Tests/Database/MultiKey_Mapper_Tests.cs
IEnumerable <DocumentTermsData> docs = tfIdfEstimator.Storage.DocumentTermsColl.Find(d => d.Terms.Select(z => z.Term).Any(x => Compare(documentTermsData1.Terms, x)));
//IEnumerable<DocumentTermsData> docs = tfIdfEstimator.Storage.DocumentTermsColl.FindAll();//.ToList();//optimize it!!! - 2020-12-22T10:04:31
// Get similarity scores of all those documents.
var docScores = new List <DocumentSimilarityScoreData>();
foreach (DocumentTermsData documentTermsData in docs)
{
if (documentTermsData1.Document == documentTermsData.Document)
{
continue;//skip it
}
docScores.Add(new DocumentSimilarityScoreData()
{
Document = documentTermsData.Document,
Score = GetDocumentSimilarityExt(documentTermsData1, documentTermsData.Document, tfIdfEstimator)
});
}
//Order them by their scores descending and return top N of them.
docScores.OrderByDescending(x => x.Score);
return(docScores.GetRange(0, numberOfDocuments));
}
public static double GetDocumentSimilarityExt(string documentId1, string documentId2, TfIdfEstimatorExt tfIdfEstimator)
{
DocumentTermsData documentTermsData1 = tfIdfEstimator.Storage.DocumentTermsColl.FindOne(x => x.Document == documentId1);
return(GetDocumentSimilarityExt(documentTermsData1, documentId2, tfIdfEstimator));
}
