Open goern opened 4 years ago
Text base analysis of the issue's body content could be used for the relevant label addition.
@GiorgosKarantonis @saisankargochhayat @xtuchyna
I've been working on the laptop and the bootcamp so far... I'll make sure to have some suggestions by tomorrow!
EDIT: Some pretty straightforward issues I see so far have to do with the language modeling and the overall network structure. Tomorrow I'll take a look on the dataset as well and see how we can augment it.
Regarding the language modeling, in the notebook they use the Keras Embedding layer which to me doesn't look like a very good option, especially for the dataset we deal with, since posted issues are usually small in size with high variance.
Regarding the network structure, we could try increasing the number of recurrent layers (depending on the size of the dataset we'll end up using) and test it with LSTMs too just for completeness.
Again, these are just straightforward suggestions and if they don't give good results I could look for a good paper to implement or come up with something more complex.
Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale
.
Stale issues rot after an additional 30d of inactivity and eventually close.
If this issue is safe to close now please do so with /close
.
/lifecycle stale
/lifecycle frozen this project would be also reviewed for this use case https://github.com/aicoe-aiops/github-labeler
Is your feature request related to a problem? Please describe. This bot will use a back end service to predict a set of labels for a given issue or pull request.
Describe the solution you'd like one back end service should expose a ml model for inference
Describe alternatives you've considered n/a
Additional context n/a