private void WriteHeader(WordModel m)
{
Writer.Write(m.Words);
Writer.Write(' ');
Writer.Write(m.Size);
Writer.Write('\n');
}
public void Write(WordModel m)
{
WriteHeader(m);
foreach (var wv in m.Vectors)
{
WriteWordVector(wv);
}
}
public void Write(WordModel m)
{
//Write the header
WriteHeader(m);
//Write the vectors
foreach (var wv in m.Vectors)
{
WriteWordVector(wv);
}
}
public static float[] GetParagraphVector(this WordModel model, params SentenceItem[] sentences)
{
if (model == null)
{
throw new ArgumentNullException(nameof(model));
}
if (sentences == null)
{
throw new ArgumentNullException(nameof(sentences));
}
if (sentences.Length == 0)
{
throw new ArgumentException("Value cannot be an empty collection.", nameof(sentences));
}
ConcurrentBag <WordVector> words = new ConcurrentBag <WordVector>();
Parallel.ForEach(sentences.SelectMany(item => item.Words),
word =>
{
if (!string.IsNullOrEmpty(word.Text))
{
var result = model.Find(word.Text);
if (result != null)
{
words.Add(result);
return;
}
}
if (!string.IsNullOrEmpty(word.Raw) && word.Raw != word.Text)
{
var result = model.Find(word.Raw);
if (result != null)
{
words.Add(result);
}
}
});
var arrays = words.ToArray();
if (arrays.Length == 0)
{
return(new float[model.Size]);
}
return(arrays.Average());
}
public static IEnumerable <WordDistance> Nearest(this WordModel model, string word)
{
var vector = model.GetByWord(word);
if (vector == null)
{
throw new ArgumentException($"cannot find word '{word}'");
}
return(model.Vectors.AsParallel()
.Select(x => new WordDistance(x.Word, x.Vector.Distance(vector.Vector)))
.OrderBy(x => x.Distance)
.Where(x => x.Word != word));
}
public static double Distance(this WordModel model, string word1, string word2)
{
var vector1 = model.GetByWord(word1);
var vector2 = model.GetByWord(word2);
if (vector1 == null)
{
throw new ArgumentException($"cannot find word1 '{word1}'");
}
if (vector2 == null)
{
throw new ArgumentException($"cannot find word2 '{word2}'");
}
return(vector1.Vector.Distance(vector2.Vector));
}
public static WordVector NearestSingle(this WordModel model, float[] vector)
{
return(model.Vectors.OrderBy(x => x.Vector.Distance(vector)).First());
}
public static IEnumerable <WordVector> Nearest(this WordModel model, float[] vector)
{
return(model.Vectors.OrderBy(x => x.Vector.Distance(vector)));
}
public static WordVector GetByWord(this WordModel model, string word)
{
return(model.Vectors.FirstOrDefault(x => x.Word == word));
}
private void WriteHeader(WordModel m)
{
WriteString($"{m.Words} {m.Size}\n");
}
