public static Task <double[, ]> VectorizeDocumentsTFIDF(IEnumerable <string> documents)
{
return(Task.Run(() =>
{
try
{
using (Py.GIL())
{
dynamic sklear_feature_extraction_text = Py.Import("sklearn.feature_extraction.text");
dynamic tfidfVectorizer = sklear_feature_extraction_text.TfidfVectorizer();
dynamic countMatrixObject = tfidfVectorizer.fit_transform(documents);
PyList countMatrix = PyList.AsList(countMatrixObject.toarray());
var matrix = (double[][])countMatrix.AsManagedObject(typeof(double[][]));
return ConvertMatrix(matrix);
}
}
catch (PythonException e)
{
if (e.Message.Contains("NoneType"))
{
return new double[0, 0];
}
else
{
throw e;
}
}
}));
}
public void TestListAsListType()
{
var l = new PyList();
PyList t = PyList.AsList(l);
Assert.IsNotNull(t);
Assert.IsInstanceOf(typeof(PyList), t);
}
private PyList GetSysPaths()
{
// set sys paths
PyObject sys = PythonEngine.ImportModule("sys");
PyObject sysPathsObj = sys.GetAttr("path");
return(PyList.AsList(sysPathsObj));
}
public void TestStringAsListType()
{
var i = new PyInt(5);
PyList t = null;
var ex = Assert.Throws <PythonException>(() => t = PyList.AsList(i));
Assert.AreEqual("'int' object is not iterable", ex.Message);
Assert.IsNull(t);
}
private static double[,] ConvertSimilarityPythonListToArray(PyList similarityList)
{
var length = similarityList.Length();
var similarities = new double[length, length];
for (int i = 0; i < length; i++)
{
var columnRows = PyList.AsList(similarityList[i]);
var row = (double[])columnRows.AsManagedObject(typeof(double[]));
for (int j = 0; j < length; j++)
{
similarities[i, j] = row[j];
}
}
return(similarities);
}
public bool TryDecode <T>(PyObject pyObj, out T value)
{
if (!PyIter.IsIterable(pyObj))
{
value = default;
return(false);
}
using (var pyList = PyList.AsList(pyObj))
{
if (typeof(T).IsGenericType)
{
value = pyList.ToList <T>();
}
else
{
value = (T)pyList.ToList();
}
return(true);
}
}
public static Task <double[, ]> FindSimilarities(string stringifiedMatrix)
{
return(Task.Run(() =>
{
using (Py.GIL())
{
dynamic pairwise = Py.Import("sklearn.metrics.pairwise");
dynamic cosine_similarity = pairwise.cosine_similarity;
dynamic np = Py.Import("numpy");
dynamic matrix = np.array(np.mat(stringifiedMatrix));
PyDict locals = new PyDict();
locals.SetItem("X", matrix);
locals.SetItem("cosine_similarity", cosine_similarity);
dynamic cosineSimilarityObj = PythonEngine.Eval("cosine_similarity(X, X)", null, locals.Handle);
PyList cosineSimilarity = PyList.AsList(cosineSimilarityObj);
return ConvertSimilarityPythonListToArray(cosineSimilarity);
}
}));
}
