Hi Prism Team,
i did not find an responsible contact so i post my request here.
Currently, we use "Latent Dirichlet Allocation" to identify semantic topics and relevant keywords in texts. Our results look very similar to your highlighted texts. I'm hardly looking for real user annotated data to verify my machine learning results. If you wish, i send you our latest paper explaining our methods and send a link to our software for demonstration. I would like to start a little project with you to evaluate and publish my machine learning results compared to your annotated data.
Kind regards,
Robin