In order to use spicedham to classify incoming feedback, we need to train the classifiers. We want to train it to denote spam/ham as well as abuse/non-abuse. This bug covers: 1. building the training sets in a development environment 2. serializing that training data into Django fixtures 3. checking in those fixtures into the repo 4. adjusting the deployment script to load those fixtures on deploy Note: This is purely for the purposes of integrating the spicedham prototype. We're not trying to design infrastructure that will be used outside of this specific need. After we figure out how things should work, it's expected we'll wipe away all this stuff and "start over".
Grabbing this to work on now.
Assignee: nobody → willkg
Status: NEW → ASSIGNED
Most of this except the training data and the deploy script fix in a PR: https://github.com/mozilla/fjord/pull/346 Landed in master: https://github.com/mozilla/fjord/commit/0c4221368e7d0212127d5572783004c982f45b11
Training data and deploy script fix in pr: https://github.com/mozilla/fjord/pull/348 Landed in master: https://github.com/mozilla/fjord/commit/78a2d630f785fc0200218c8d45055742cc936dc1
Pushed this to prod last night.
Status: ASSIGNED → RESOLVED
Last Resolved: 3 years ago
Resolution: --- → FIXED
Whiteboard: u=dev c=spam p= s=input.2014q3 → u=dev c=spam p=1 s=input.2014q3
You need to log in before you can comment on or make changes to this bug.