How to use the "class torch.nn.CTCLoss"

I want to use the CTCLoss in my code and I read the docs carefully. But I still don’t understand what each of the parameters means. For example, I want to deal with a video, where a person said a sentence containing several words,how do I set the parameters in this function???