I am developing an online expert system in Python that generates psychological reports (based on psychological tests). The report consists of sentences (ie. personality description), constructed by elementary statements in a simple grammatical structure.
As an addition, I would like to test if a NN solution could improve the quality of the report. Users will be asked to somehow judge each sentence whether it is true or false (or perhaps rate them on a numeric scale).
I would like to collect user ratings and use it for the improvement of the report, e.g. sentences which were downvoted by users should appear with a lower probability and the best ones should be more frequent.
Could anyone advise me if neural networks are a good fit for this particular problem? If yes, how can I choose the NN architecture?