I am working with a homegrown language that distills down ideas into a short sentence. In short, it takes several paragraphs and ‘translates’ them into a few tokens.
I’m wondering if there is a good NMT approach that might work for this. I was even thinking of just using Bertmulticlassification. But I am also trying to see if NMT is appropriate. I know this introduces fairly serious alignment issues, but was hoping someone could point me in a direction to start.