Adding labeled utterances for reference in Bulk Labeling Tool

[oppasource]

Thanks for the amazing work!! Loved it.

Bulk Labeling Tool currently only takes unlabeled utterances. Is it possible to visualize few already labeled utterances as well (possibly with different color) in the same semantic space. This will allow users to visualize separation of clusters more clearly. For example there might be confusion of considering a sub-cluster or a bigger super-cluster while labeling. Colored dots(few labeled samples) can help guide the manual cluster selection process.

Also a drop down menu for all available intent that are already present in the current labeled version would go a long way I feel.

[sara-tagger]

Thanks for raising this issue, @yennycheung will get back to you about it soon✨

Please also check out the docs and the forum in case your issue was raised there too 🤗

[koaning]

So, technically, we already support labelled utterances but you need to pass extra settings to the human-learn code. As you can see in the docs here, it's totally allowed. You probably need to add something like below to the InteractiveCharts call.

clf = InteractiveCharts(df, labels="whatever-label-name")

[koaning]

A drop-down menu might work well here, but I should start by admitting that there is a reason why the current state of the project is a Jupyter notebook as opposed to a python package; it is still very much work in progress. One thing that I'm also exploring is if it makes sense to add a model in the mix for active learning. By releasing it as a notebook my hope is that people feel free to hack around themselves as well and that they might let me know what works well.

[oppasource]

Thanks taking time out for your helpful response.