Text this: Recognizing American Sign Language gestures efficiently and accurately using a hybrid transformer model