The rating system is a human in the loop: me. So it was my judgement call as to what the title accuracy entailed. My goal was to provide tools so that other interested parties would be able to make their own assessment, and that they could check my logs to verify. The logs are all included in that folder.
Hi jbash, I dove a little deeper on my title accuracy system here: https://github.com/dhealy05/semen_and_semantics/blob/main/analysis_results/title_accuracy_logs/title_accuracy_readme.md but didn’t account for it when I transferred the readme to the LessWrong format.
The rating system is a human in the loop: me. So it was my judgement call as to what the title accuracy entailed. My goal was to provide tools so that other interested parties would be able to make their own assessment, and that they could check my logs to verify. The logs are all included in that folder.
For example, to rate the videos surfaced in https://github.com/dhealy05/semen_and_semantics/blob/main/analysis_results/title_accuracy_logs/incest_title_accuracy.json I visited each URL and ranked on the 1-5 scale. “Cum in panties step sister” did not seem to involve a step sister, so I gave it a 1: SEO effect. “Kinky Family—Home alone with slutty stepsis” does indeed seem to involve a stepsis oriented plot, so I gave it a 5: no SEO effect.