Feature Request - Speaker identification with limited words

Hi,
Recognizer constructor doesn't accept word list when we provide a speaker model, is it possible to add a constructor which can accept speaker models and custom words list.
for example.
`rec = KaldiRecognizer(model, spk_model, wf.getframerate(), '["oh one two three four five six seven eight nine zero", "[unk]"]')`
currently I cannot use both speaker identification and custom words, I have to create two recognizer objects first get the text and then verify the speaker in second one.
I want to use limited words only as they increase accuracy and are almost instant on my rpi.