List view

• 0.5 - Integrate with Kesh

  Once a model has been created, new questions must be tested against it to see whether they should be closed or not. This will require integration with Kesh. Kesh will feed question information to Nidaba for analysis, and Nidaba can then send the results back to Kesh for sending to other projects (RABBIT, sopython-site, etc)

  No due date
  0% complete0 open0 closed

• 0.4 - Learn initial model

  Once features have been chosen and a ML process established, it will be necessary to create the initial model. This initial model will be used to predict problems in the future, and will use all of the information we have to date.

  No due date
  0% complete0 open0 closed

• 0.3 - Feature discovery

  Once ML can be undertaken, the quality and necessity of a variety of features will be better understood. At this stage experimentation should be undertaken so a decent set of features can be found to move forward with.

  No due date
  0% complete0 open0 closed

• 0.2 - Machine Learning process

  Once information can be extracted as features, it will be necessary to use these features in a ML process that can then be used to predict future features.

  No due date

  •0/1 issues closed
  0% complete1 open0 closed

• 0.1 - Feature extraction code

  The first sub-part of the Nidaba project will be analysing whether questions should be closed or not. We will not be analysing duplicate closure (yet) but will instead be classifying questions on whether they are: - 0 - fine (and hence remain unclosed) - 1 - Too broad - 2 - Primarily opinion based - 3 - Unclear - 4 - Off-topic (possibly the OT reasons will be expanded to 5, 6, 7...) Code should be written that can be used to extract various features from SO questions and answers. Ideally the code should be modular enough that it can be easily modified when Nidaba is running as the input/output might be different later on.

  No due date

  •4/9 issues closed
  44% complete5 open4 closed