You are reading a single comment by @Emyr and its replies. Click here to read the full conversation.
  • They could do with tying their new-products-review process to Bayesian analysis of the feed.

    i.e. Based on log data, 70% of reviewed submissions with keywords x,y and z in the blurb were declined. 20% of reviewed submissions with keywords x,y,z and denial in the blurb were declined etc etc.

    Based on the keywords in this blurb, weighted by the strength of data held, how likely is it that we should block this book? Tweak the threshold so that the manual review workload fits the number of reviewers you employ. Increase number of reviewers if the trend of user reports for unreviewed items goes up.

    Or, y'know, MapReduce that shit using Hadoop.

About

Avatar for Emyr @Emyr started