static void Main2(string[] args)
{
//Ensures that we are consistent against culture-specific number formating, etc...
CultureInfo culture = CultureInfo.CreateSpecificCulture("en-US");
CultureInfo.DefaultThreadCurrentCulture = culture;
CultureInfo.DefaultThreadCurrentUICulture = culture;
Thread.CurrentThread.CurrentCulture = culture;
Thread.CurrentThread.CurrentUICulture = culture;
if (!File.Exists(@"cheeseDisease.bin"))
{
Console.WriteLine("Please train the model first using the console version of fastText and the data supplied in the SampleData folder");
Console.WriteLine("fastText.exe supervised -input cheeseDisease.txt -output cheeseDisease");
return;
}
Console.WriteLine("Loading model, please wait");
var fastTextModel = new fastText(@"cheeseDisease.bin");
Console.WriteLine("... done!");
var tests = GetTestData();
int correct = 0, fail = 0, noLabel = 0;
foreach (var test in tests)
{
var label = fastTextModel.GetPrediction(test.Text, 1).First().label.Replace("__label__", "").Replace("__", "");
if (label == "n/a")
{
label = fastTextModel.GetPrediction(test.Text, 1).First().label.Replace("__label__", "").Replace("__", "");
}
Console.WriteLine($"{test.Text} -> P:{label} / C:{test.Label}");
correct += (label == test.Label) ? 1 : 0;
fail += (label == test.Label) ? 0 : 1;
noLabel += (label == "n/a") ? 1 : 0;
}
Console.WriteLine($"Summary: {correct} correctly labeled, {fail-noLabel} mislabed, {noLabel} no labels found");
Console.WriteLine("Press any key to finish!");
Console.Read();
fastText.Release();
}
public static List <SentenceInterpretation> IdentifyEntities(fastText fastTextModel, string text, int numberOfPossibilities)
{
var paths = new List <SentenceInterpretation>();
var words = text.Split(new char[] { ' ' }, StringSplitOptions.RemoveEmptyEntries).ToList(); //IMPROVE TOKENIZATION HERE
if (words.Count == 0)
{
return(paths);
}
var tokens = new List <List <Token> >();
foreach (var w in words)
{
var pred = fastTextModel.GetPrediction(w, 5);
if (pred.Count == 0)
{
pred.Add(new prediction()
{
label = "S_O", intensity = 1
});
}
tokens.Add(pred.Select(p => new Token(w, (p.label.Contains("_") ? p.label : "S_" + p.label), p.intensity)).ToList());
}
var graph = new SentenceGraph();
tokens.ForEach(t => graph.AddVertexRange(t));
//tokens.ForEach(p => Console.WriteLine("\t" + string.Join(" ", p.Select(t => t.Label + "[" + t.Intensity.ToString("0.0") + "]"))));
for (int i = 0; i < (words.Count - 1); i++)
{
double maxIntensityC = tokens[i].Max(p => p.Intensity);
double maxIntensityN = tokens[i + 1].Max(p => p.Intensity);
foreach (var source in tokens[i])
{
foreach (var dest in tokens[i + 1])
{
string clabel = source.Label;
string nlabel = dest.Label;
if (!source.Label.Contains("_"))
{
clabel = SentenceGraph.SingleTag + "_" + source.Label;
}
if (!dest.Label.Contains("_"))
{
nlabel = SentenceGraph.SingleTag + "_" + dest.Label;
}
if (SentenceGraph.IsTransitionAllowed(clabel, nlabel))
{
double probability = (source.Intensity / maxIntensityC) * (dest.Intensity / maxIntensityN);
if (source.Intensity < 0 || dest.Intensity < 0)
{
probability = 0;
}
graph.AddEdge(new SentenceEdge(source, dest, clabel, nlabel, probability));
//Console.WriteLine($"\tFound {source.Value}[{clabel}] -> {dest.Value}[{nlabel}] with probablity {probability} and intensities {source.Intensity} and {dest.Intensity}");
}
}
}
}
var BoS = new Token("__BEGIN__", "", 1);
var EoS = new Token("__END__", "", 1);
graph.AddVertex(BoS); graph.AddVertex(EoS);
foreach (var t in tokens.First())
{
string tlabel = t.Label;
if (!t.Label.Contains("_"))
{
tlabel = SentenceGraph.SingleTag + "_" + t.Label;
}
if (SentenceGraph.IsTransitionAllowed("S_O", tlabel))
{
graph.AddEdge(new SentenceEdge(BoS, t, "S_O", tlabel, 1.0));
}
}
foreach (var t in tokens.Last())
{
string tlabel = t.Label;
if (!t.Label.Contains("_"))
{
tlabel = SentenceGraph.SingleTag + "_" + t.Label;
}
if (SentenceGraph.IsTransitionAllowed(tlabel, "S_O"))
{
graph.AddEdge(new SentenceEdge(t, EoS, tlabel, "S_O", 1.0));
}
}
paths = graph.GetAllPossibleSentenceInterpretations(BoS, EoS, numberOfPossibilities);
//paths.ForEach(p => Console.WriteLine("Probability: " + p.Probability.ToString("0.00") + "\t" + string.Join(" ", p.Sentence.Select(n => string.Join(" ", n.Tokens.Select(t => t.Value + "[" + t.Label + "]" ) ) ))));
return(paths);
}
