I'm currently using google speech to text api to transcribe audio in real time (police scanner audio dispatches). The audio quality isn't great and I've been putting in key words to try to help train it. Is there a way to either use google, amazon aws, IBM Watson, to create a model based on past audio dispatches where I can manually type in what was said to help train it? It seems like putting in key words won't really cut it. Any other suggestions to help make it more accurate?