static void Main3(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;
string model = @"C:\IW\VS\PROJECTS\BOSON\Boson.Testing.CreateLuisApp\bin\x64\Debug\training\ner-caseinsensitive-fasttext";
var fastTextModel = new fastText(model + ".bin");
//var tags = File.ReadAllLines(model + @".vec").Skip(2).Where(l => l.StartsWith("nertag")).Select(l => l.Split(' ').First()).ToList();
while (true)
{
var text = Console.ReadLine();
var paths = FastTextNER.SentenceGraph.IdentifyEntities(fastTextModel, text, 5);
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 + "]"))))));
//var pred = fastTextModel.GetPrediction(text, 5);
//pred.ForEach(p => Console.WriteLine(p.label + "[" + p.intensity.ToString("0.00") + "]"));
//foreach (var t in tags)
//{
// Console.WriteLine(t + " [" + fastTextModel.GetWordSimilarity(text, t).ToString("0.00") + "]");
//}
}
return;
}
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();
}
static void Main(string[] args)
{
string model = @"C:\BigData\NLPmodels\FastText\aviation-caseinsensitive";
if (args.Length > 0)
{
model = args[0];
}
//var words = File.ReadAllLines(model + @".vec").Skip(2).Select((l) => new WordAndSim() { word = l.Split(' ').First() }).ToList();
//words.RemoveAll(x => x.word.Length < 6);
var fastTextModel = new fastText(model + @".bin");
var words = fastTextModel.GetWords();
//var utterances = File.ReadAllLines(@"C:\stanford-nlp\classifier_training\test.tsv").Select(l => new IntentExample() { Intent = l.Split('\t').First(), Example = l.Split('\t').Last().ToLowerInvariant() }).ToList();
while (true)
{
Console.Write("\nWord: "); var w1 = Console.ReadLine();
Console.WriteLine("\nMost similar:");
fastTextModel.GetMostSimilar(w1, 20).ForEach(ws => Console.WriteLine("\t" + ws.Item1.PadLeft(15) + " " + ws.Item2.ToString("0.00")));
Console.WriteLine("\nLeast similar:");
fastTextModel.GetLeastSimilar(w1, 5).ForEach(ws => Console.WriteLine("\t" + ws.Item1.PadLeft(15) + " " + ws.Item2.ToString("0.00")));
//Console.WriteLine();
//Console.Write("Parent: "); var parent = Console.ReadLine();
//double[] averageDiff = new double[fastTextModel.GetVectorSize()];
//double count = 0;
//while(true)
//{
// Console.Write("Child : "); var child = Console.ReadLine();
// if(string.IsNullOrWhiteSpace(child)) { break; }
// var tmpdiff= fastTextModel.GetWordDifference(child, parent);
// averageDiff = fastText.Add(averageDiff, tmpdiff);
//}
//fastText.Multiply(averageDiff, 1 / count);
//Console.Write("New Parent : "); var newParent = Console.ReadLine();
//Console.Write("New Child : "); var newChild = Console.ReadLine();
////var diff = fastTextModel.GetWordDifference(child, parent);
//var newDiff = fastTextModel.GetWordDifference(newChild,newParent);
//Console.WriteLine(string.Join("; ", averageDiff.Select(a => a.ToString("0.00"))));
//Console.WriteLine(string.Join("; ", newDiff.Select(a => a.ToString("0.00"))));
//Console.WriteLine("Similarity between diff vectors:" + fastText.CalculateCosineSimilarity(averageDiff, newDiff));
//if(string.IsNullOrWhiteSpace(parent) || string.IsNullOrWhiteSpace(newParent)) { break; }
//var newChildVector = fastText.Add(fastTextModel.GetWordVector(newParent), averageDiff);
//Console.WriteLine("Similarity between projected new parent and new child: " + fastTextModel.GetWordSimilarity(newChild, newChildVector));
//foreach (var w in words)
//{
// w.sim = fastTextModel.GetWordSimilarity(w.word, newChildVector);
//}
//words.Sort((a, b) => b.sim.CompareTo(a.sim));
//Console.WriteLine("Most similar:");
//foreach (var w in words.Take(20))
//{
// Console.WriteLine($"\t{w.word} -> {w.sim}");
//}
////words.Reverse();
////Console.WriteLine();
////Console.WriteLine("Least similar:");
////foreach (var w in words.Take(5))
////{
//// Console.WriteLine($"\t{w.word} -> {w.sim}");
////}
//Console.WriteLine();
}
fastText.Release();
//while (true)
//{
// Console.Write("First word: "); var w1 = Console.ReadLine();
// Console.Write("Second word: "); var w2 = Console.ReadLine();
// Console.WriteLine($"Similarity between {w1} and {w2} is {fastText.GetWordSimilarity(w1, w2)}");
// Console.WriteLine();
//}
}
public static List <SentenceInterpretation> IdentifyEntities(fastText fastTextModel, string text, int numberOfPossibilities)
{
var paths = new List <SentenceInterpretation>();
var words = text.Split(new char[] { ' ' }, StringSplitOptions.RemoveEmptyEntries).ToList(); //IMPROVE TOKENIZATION HERE
if (words.Count == 0)
{
return(paths);
}
var tokens = new List <List <Token> >();
foreach (var w in words)
{
var pred = fastTextModel.GetPrediction(w, 5);
if (pred.Count == 0)
{
pred.Add(new prediction()
{
label = "S_O", intensity = 1
});
}
tokens.Add(pred.Select(p => new Token(w, (p.label.Contains("_") ? p.label : "S_" + p.label), p.intensity)).ToList());
}
var graph = new SentenceGraph();
tokens.ForEach(t => graph.AddVertexRange(t));
//tokens.ForEach(p => Console.WriteLine("\t" + string.Join(" ", p.Select(t => t.Label + "[" + t.Intensity.ToString("0.0") + "]"))));
for (int i = 0; i < (words.Count - 1); i++)
{
double maxIntensityC = tokens[i].Max(p => p.Intensity);
double maxIntensityN = tokens[i + 1].Max(p => p.Intensity);
foreach (var source in tokens[i])
{
foreach (var dest in tokens[i + 1])
{
string clabel = source.Label;
string nlabel = dest.Label;
if (!source.Label.Contains("_"))
{
clabel = SentenceGraph.SingleTag + "_" + source.Label;
}
if (!dest.Label.Contains("_"))
{
nlabel = SentenceGraph.SingleTag + "_" + dest.Label;
}
if (SentenceGraph.IsTransitionAllowed(clabel, nlabel))
{
double probability = (source.Intensity / maxIntensityC) * (dest.Intensity / maxIntensityN);
if (source.Intensity < 0 || dest.Intensity < 0)
{
probability = 0;
}
graph.AddEdge(new SentenceEdge(source, dest, clabel, nlabel, probability));
//Console.WriteLine($"\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);
}
