Consider using (parts of) filename as speech recognizer type

[greenw0lf]

In the evaluations I conducted, I would usually transcribe a specific subset using different models, then run the script with the results from all the models on that one subset. It would be an idea to maybe change the kind flag in such a way that it toggles using the name of the file as the speech recognizer used. But, I am not sure how well it can be integrated with the category.