public NgramModel(int ngramSize, TSeq[] sequences, Func <TSeq, IEnumerable <TItem> > itemsSelector, Direction dir, INgramModelSmoother <TSeq, TItem> smoother)
{
_ngramSize = ngramSize;
_dir = dir;
_smoother = smoother;
_ngrams = new HashSet <Ngram <TItem> >();
var cfd = new ConditionalFrequencyDistribution <Ngram <TItem>, TItem>();
foreach (TSeq seq in sequences)
{
TItem[] items = itemsSelector(seq).ToArray();
for (int i = 0; i <= items.Length - ngramSize; i++)
{
var ngram = new Ngram <TItem>(Enumerable.Range(i, _ngramSize).Select(j => items[j]));
_ngrams.Add(ngram);
Ngram <TItem> context = ngram.TakeAllExceptLast(dir);
TItem item = ngram.GetLast(dir);
cfd[context].Increment(item);
}
}
_smoother.Smooth(ngramSize, sequences, itemsSelector, dir, cfd);
}
public static NgramModel <TSeq, TItem> Train(int ngramSize, IEnumerable <TSeq> sequences, Func <TSeq, IEnumerable <TItem> > itemsSelector, Direction dir, INgramModelSmoother <TSeq, TItem> smoother)
{
return(new NgramModel <TSeq, TItem>(ngramSize, sequences.ToArray(), itemsSelector, dir, smoother));
}
public static NgramModel <TSeq, TItem> Train(int ngramSize, IEnumerable <TSeq> sequences, Func <TSeq, IEnumerable <TItem> > itemsSelector, INgramModelSmoother <TSeq, TItem> smoother)
{
return(Train(ngramSize, sequences, itemsSelector, Direction.LeftToRight, smoother));
}
