public ArchiveDocumentType? FindApproximation(string line)
{
var item = new ApproximationItem(ArchiveDocumentType.Other, line);
var foundProximity = 0.5;
ApproximationItem foundItem = null;
foreach (var approximationItem in Basis)
{
var proximity = item.FindProximity(approximationItem);
if (proximity > foundProximity)
{
foundProximity = proximity;
foundItem = approximationItem;
}
}
return foundItem != null ? foundItem.Type : default(ArchiveDocumentType?);
}
public double FindProximity(ApproximationItem item)
{
var oneLetterOccurrenceCounter = 0;
var oneLetterOccurrenceTotal = item.OneLetterOccurrence.Count;
foreach (var occurrence in item.OneLetterOccurrence)
{
if (!OneLetterOccurrence.ContainsKey(occurrence.Key)) continue;
var foundOccurrence = OneLetterOccurrence.First(x => x.Key == occurrence.Key);
var value = (foundOccurrence.Value <= occurrence.Value)
? foundOccurrence.Value
: 2*occurrence.Value - foundOccurrence.Value;
oneLetterOccurrenceCounter += value;
}
var twoLetterOccurrenceCounter = 0;
var twoLetterOccurrenceTotal = item.TwoLetterOccurrence.Count;
foreach (var occurrence in item.TwoLetterOccurrence)
{
if (!TwoLetterOccurrence.ContainsKey(occurrence.Key)) continue;
var foundOccurrence = TwoLetterOccurrence.First(x => x.Key == occurrence.Key);
var value = (foundOccurrence.Value <= occurrence.Value)
? foundOccurrence.Value
: 2 * occurrence.Value - foundOccurrence.Value;
twoLetterOccurrenceCounter += value;
}
var threeLetterOccurrenceCounter = 0;
var threeLetterOccurrenceTotal = item.ThreeLetterOccurrence.Count;
foreach (var occurrence in item.ThreeLetterOccurrence)
{
if (!ThreeLetterOccurrence.ContainsKey(occurrence.Key)) continue;
var foundOccurrence = ThreeLetterOccurrence.First(x => x.Key == occurrence.Key);
var value = (foundOccurrence.Value <= occurrence.Value)
? foundOccurrence.Value
: 2 * occurrence.Value - foundOccurrence.Value;
threeLetterOccurrenceCounter += value;
}
return (double)(oneLetterOccurrenceCounter + 2*twoLetterOccurrenceCounter + 3*threeLetterOccurrenceCounter)/
(double)(oneLetterOccurrenceTotal + 2*twoLetterOccurrenceTotal + 3*threeLetterOccurrenceTotal);
}
