public void DeploySettings(IPlaneSettings settings, ToolkitExperimentNotes _notes, ILogBuilder logger)
{
DeploySettingsBase(_notes);
PlanesMethodSettings mainSettings = (PlanesMethodSettings)settings;
EntityMethod.DeploySettings(mainSettings.entityMethod, notes, logger);
CorpusMethod.DeploySettings(mainSettings.corpusMethod, notes, logger);
VectorMethod.DeploySettings(mainSettings.vectorMethod, notes, logger);
FeatureMethod.DeploySettings(mainSettings.featureMethod, notes, logger);
CacheProvider.Deploy(new System.IO.DirectoryInfo(mainSettings.cachePath));
EntityMethod.CacheProvider = CacheProvider;
CorpusMethod.CacheProvider = CacheProvider;
VectorMethod.CacheProvider = CacheProvider;
FeatureMethod.CacheProvider = CacheProvider;
CloseDeploySettingsBase();
}
/// <summary>
/// Gets default configuration
/// </summary>
/// <returns></returns>
public static PlanesMethodSettings GetDefaultSettings()
{
PlanesMethodSettings output = new PlanesMethodSettings();
output.entityMethod.instructions.Add(DocumentRenderInstruction.GetDescriptionInstruction());
output.entityMethod.instructions.Add(DocumentRenderInstruction.GetTitleInstruction());
output.entityMethod.instructions.Add(DocumentRenderInstruction.GetBodyTextInstruction());
// output.entityMethod.blenderOptions = DocumentBlenderFunctionOptions.binaryAggregation | DocumentBlenderFunctionOptions.pageLevel;
output.entityMethod.filterFunctionName = ""; // nameof(DocumentEntropyFunction);
output.entityMethod.filterLimit = 5;
output.corpusMethod.stemmer = nameof(EnglishStemmer);
output.corpusMethod.tokenizer = nameof(TokenizerBasic);
output.corpusMethod.transliterationRuleSetId = "";
#region PREPARE Weighting model
var weightModel = new FeatureWeightModel();
weightModel.LocalFunction = new Weighting.Local.TermFrequencyFunction();
var globalFactor = new FeatureWeightFactor();
globalFactor.Settings.functionName = nameof(IDFElement);
weightModel.GlobalFactors.Add(globalFactor);
output.corpusMethod.WeightModel = weightModel;
#endregion
var featureFilter = new FeatureFilter();
featureFilter.limit = 8000;
//featureFilter.
//featureFilter.functionSettings = new GlobalFunctionSettings();
//featureFilter.functionSettings.functionName = nameof(CollectionTDPElement);
//featureFilter.functionSettings.weight = 1.0;
//featureFilter.functionSettings.flags.Add(Weighting.Metrics.TDPFactor.chi.ToString());
output.corpusMethod.filter = featureFilter;
/*
* output.vectorMethod.constructor = new Feature.Settings.FeatureVectorConstructorSettings();
* dimensionSpecification dimSpec = new dimensionSpecification();
* dimSpec.functionName = nameof(CosineSimilarityFunction);
* dimSpec.type = FeatureVectorDimensionType.similarityFunction;
* output.vectorMethod.constructor.labelDimensions.Add(dimSpec);
*/
//output.vectorMethod.constructor = new Feature.Settings.FeatureVectorConstructorSettings();
//dimensionSpecification dimSpec = new dimensionSpecification();
////dimSpec.functionName = nameof(CosineSimilarityFunction);
//dimSpec.type = FeatureVectorDimensionType.directTermWeight;
//output.vectorMethod.constructor.featureDimensions.Add(dimSpec);
output.featureMethod.classifierSettings.type = Classifiers.ClassifierType.multiClassSVM;
output.featureMethod.classifierSettings.lossFunctionForTraining = Accord.MachineLearning.VectorMachines.Learning.Loss.L2;
/*
* output.featureMethod.classifierSettings.type = Classifiers.ClassifierType.kNearestNeighbors;
* output.featureMethod.classifierSettings.lossFunctionForTraining = Accord.MachineLearning.VectorMachines.Learning.Loss.L2;
* output.featureMethod.classifierSettings.kNN_k = 4;
*/
return(output);
}
