Suppose one is building a classifier that:

  1. takes as input the e-mail body text
  2. returns true or false if person X should be included in the address list

To build this classifier we have historical data of e-mail correspondence from a community of users Y.


  1. Persons John, Jill and Sally are members of Y
  2. Persons John and Jill are married
  3. Persons John, Jill and Sally are mutual friends
  4. Persons John and Sally are having an affair unbeknownst to Jill

The situation we want to avoid is:

  1. Every time Jill writes an intimate e-mail to John
  2. The classifier keeps recommending that she include Sally on the address list

So the question is how does one partition the training data to avoid such "contamination" of signals?

  • $\begingroup$ that's one classifier per person? maybe you can add more details on how you are intending on setting this up. otherwise it's not clear how to answer the question. $\endgroup$ – oW_ May 24 at 18:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.