Build a Machine Learning Classification dataset from your Gmodstore tickets, using the Gmodstore Tag system.
The script used the Gmodstore API to pull all the given tickets for a provided product, and breaks their opening message up into a folder structure based on the ticket's labels. It follows the structure required by liner.ai's classification model, which is what this script was made to satisfy. But there's no reason you couldn't import this data to another tool/package.
If you use liner.ai, you can export your model as a TensorFlow. You can then use your model in something like NodeJS to automatically classify new Gmodstore Tickets and provide automated responses, which is the reason this even exists.
- Clone this repo.
- Run
npm install
. - Copy the
.env.copy
to.env
. - Fill in the 2 values for the
.env
. - Run
npm run start
. - Allow the script to pull down the data from the API and format it.
- You should now find a labelled data set in
/data
Any untagged tickets will be placed in a /unlabelled
label. You are then free to label those yourself. It is not suggested to include the unlabelled
label set in your training data, as it can warp your results. I suggest labelling your tickets
on GmodStore directly, because if you need to pull the data down again at a later date, you will not need to manually relabel it all.
- Creekie - Helping me understand things and being an emotional support robot.