I'm working on a AI project which generates poems in Turkish language like this project: https://github.com/schollz/poetry-generator
It's a successful project but it's a template based poem generator. You can't select a topic for poem generation. It generates poems randomly with hardcoded templates and words.
In my project I also have some templates and huge lexicon but there is a difference. User can select a topic or give a sentence to identify poem's theme. Program does word similarity analyse and select word according to these results.
I'm not finished this project yet. But I believe I will soon. At the end, I want to do an experiment/turing test in order to calculate it's success ratio. But I'm not sure which way should I follow.
My idea: I will publish an experiment announcement in my university. I won't say it's about AI or computer science. It will be like poem contest or something. I will show the human generated and ai generated poems to human subjects. They will score the poems according to their taste of art. After the first phase is finished, I will tell them some of the poems were generated by an AI. After then they will try to identify AI generated poems. I'm not sure it's a suitable turing test. What is the best practice for this one?