static void Main(string[] args)
{
var encoder = new CRFSharpWrapper.Encoder();
var options = new EncoderArgs();
options.debugLevel = 1;
options.strTemplateFileName = @"D:\NLP\template.NE"; //template file name
options.strTrainingCorpus = @"D:\NLP\training.txt"; //training corpus file name
options.strEncodedModelFileName = @"D:\NLP\ner_model"; //encoded model output file name
options.max_iter = 1000;
options.min_feature_freq = 2;
options.min_diff = 0.0001;
options.threads_num = 1;
options.C = 1.0;
options.slot_usage_rate_threshold = 0.95;
bool bRet = encoder.Learn(options);
}
static void Main(string[] args)
{
/*training*/
EncoderConsole encoder = new EncoderConsole();
EncoderArgs encoderArg = new EncoderArgs("ExampleData\\training.txt", "ExampleData\\template", "ExampleData\\model");
encoder.Train(encoderArg);
/*predixt*/
DecoderConsole decoder = new DecoderConsole("ExampleData\\model");//執行這一行就會先把model load入記憶體
string predict = File.ReadAllText("ExampleData\\test.txt");
DecoderArgs decoderArg = new DecoderArgs(predict, 0);
string result = decoder.Predict(decoderArg);
Console.WriteLine(result);
Console.ReadKey();
}
public void Run(string [] args)
{
CRFSharpWrapper.Encoder encoder = new CRFSharpWrapper.Encoder();
CRFSharpWrapper.EncoderArgs options = new EncoderArgs();
for (int i = 0; i < args.Length; i++)
{
if (args[i][0] == '-')
{
string key = args[i].Substring(1).ToLower().Trim();
string value = "";
if (key == "encode")
{
continue;
}
if (key == "debug")
{
options.debugLevel = 1;
try
{
if (i < args.Length - 1)
{
int debugLevel = int.Parse(args[i + 1]);
options.debugLevel = debugLevel;
i++;
}
}
catch (Exception) {}
}
else if (i < args.Length - 1)
{
i++;
value = args[i];
switch (key)
{
case "template":
options.strTemplateFileName = value;
break;
case "trainfile":
options.strTrainingCorpus = value;
break;
case "modelfile":
options.strEncodedModelFileName = value;
break;
case "maxiter":
options.max_iter = int.Parse(value);
break;
case "minfeafreq":
options.min_feature_freq = int.Parse(value);
break;
case "mindiff":
options.min_diff = double.Parse(value);
break;
case "thread":
options.threads_num = int.Parse(value);
break;
case "costfactor":
options.C = double.Parse(value);
break;
case "slotrate":
options.slot_usage_rate_threshold = double.Parse(value);
break;
case "hugelexmem":
options.hugeLexMemLoad = uint.Parse(value);
break;
case "retrainmodel":
options.strRetrainModelFileName = value;
break;
case "regtype":
if (value.ToLower().Trim() == "l1")
{
options.regType = CRFSharpWrapper.Encoder.REG_TYPE.L1;
}
else if (value.ToLower().Trim() == "l2")
{
options.regType = CRFSharpWrapper.Encoder.REG_TYPE.L2;
}
else
{
Console.WriteLine("Invalidated regularization type");
Usage();
return;
}
break;
default:
ConsoleColor cc = Console.ForegroundColor;
Console.ForegroundColor = ConsoleColor.Red;
Console.WriteLine("No supported {0} parameter, exit", key);
Console.ForegroundColor = cc;
Usage();
return;
}
}
else
{
ConsoleColor cc = Console.ForegroundColor;
Console.ForegroundColor = ConsoleColor.Red;
Console.WriteLine("{0} is invalidated parameter.", key);
Console.ForegroundColor = cc;
Usage();
return;
}
}
}
if (options.strTemplateFileName == null || options.strEncodedModelFileName == null || options.strTrainingCorpus == null)
{
Usage();
return;
}
if (options.threads_num <= 0)
{
options.threads_num = Environment.ProcessorCount;
}
bool bRet;
bRet = encoder.Learn(options);
}
