I have a question regarding the model adaptability in a slightly different case, I want to train a generic sentiment classifier on multiple domains (multiple datasets), without having a specific target domain, how would that fit in the current model?
Sure, you can just train a multinomial adversarial network as the P-net using multiple domains and another networks for learning domain-specific information.
Hi, thank you for releasing your work!
I have a question regarding the model adaptability in a slightly different case, I want to train a generic sentiment classifier on multiple domains (multiple datasets), without having a specific target domain, how would that fit in the current model?