nikgil / senti-bot Goto Github PK

Can be gotten from https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge/data or alternatively added yourself. This is not a single commit issue.

Below is a list of commands that should be added to the bot to be functioning. This is prone to change as time moves forward.

• Mute Channel - this is to remove the bot from a certain channel and ignore it.
• List Channels - from above, to list where it is inactive.
• Set Tolerance - change tolerance level for the test method.
• Retrain - start the training algorithm again.
• Pickle - save classifier data (potentially using NLTK library).
• Choose Kernel - ability to change kernels on the fly. This may mean training them in advance or working with #4 to train the kernel in parallel after selecting it.

Because the training method may take a substantial amount of time, it may be prudent to train it on its own thread rather than stop-the-world on startup.

One of the first kernels I want to use is from https://www.kaggle.com/jhoward/nb-svm-strong-linear-baseline as it seems to create a strong baseline for future ones.

From the Travis builds, JobLib is deprecated and should be switched out from something else within Python.

Create general test cases that all kernels should suffice before being added to master.

Possibly should be done after a baseline kernel is made for comparison #1 .

Ability to train algorithms dynamically. At the very least this can be private messaging the bot messages and stating what category they fall into.

README will need updates for newer installation process (after being dockerized) and new commands/usage as needed. This is a multi-commit issue.

• Add Travis CI badge to top of file
• Correct Python version to be a minimum of 3.6 (this may change with CI testing)