I’m very new to NLP and don’t yet fully grasp the different tasks it can be used for. Apologies for butchering the terminology.
My task requires the creation of natural language summaries using keywords/phrases as input. Is this something that can be accomplished immediately using existing models, or would it perhaps require a multi-stage approach?
I have encountered a project called keytotext by gagan3012 that summarises/generates text using only keywords/phrases as input, however the resulting text is not of suitably-high fidelity.
If I wanted to create the best possible text from my (extensive) set of keywords, how would I go about it? If the multi-stage approach is preferable, would I perhaps first perform topic modelling on the set of keywords/phrases, then use TextRank to generate importance scores, use something like keytotext to create a first-run summary before using NLU to refine the resulting text for readability?
I would be very grateful for any guidance or recommendations.