Don't start the process directly; the details of the command line are version-dependent (and, in fact, the process itself may change from version to version).
You can start training (in C#) using speechlib (the SAPI IDispatch-compatible API). Look at ISpeechRecognizer::DisplayUI.
To use SpeechLib, add
using SpeechLib;
to your code, and add a reference (via Project/Add Reference/COM) to the Microsoft Speech Object Library
to your project.
Then, to start training, you would have some code that looks like this:
static void RunTraining()
{
SpSharedRecoContext RC = new SpSharedRecoContext();
string Title = "My App's Training";
ISpeechRecognizer spRecog = RC.Recognizer;
spRecog.DisplayUI(hWnd, Title, SpeechLib.SpeechStringConstants.SpeechUserTraining, "");
}