search

Data4Democracy / internal-displacement

Studying news events and internal displacement.
43 stars 27 forks source link

Implement online learning #145

Open georgerichardson opened 7 years ago

georgerichardson commented 7 years ago

As more articles are gathered, analysed and verified by a human, it would be nice for the ML models to self update. Open to discussion on tools and best practices for this.

domingohui commented 7 years ago

What are the ML models used for? I must have lost track in the slack convo...

georgerichardson commented 7 years ago

The ML I'm thinking of for this particular issue are to tag the articles as about displacement/not about displacement and then as disaster/conflict/other as the reason for displacement

oskarsinger commented 6 years ago

I have an extensive online learning background. Would be interested in details of the problem. Sounds like it may be better to just retrain periodically because this sounds like a supervised learning problem, but it depends on more specifics.

emtseng commented 6 years ago

Agreed @oskarsinger. Details pending, of course, but I imagine it may be useful to investigate and then set a threshold for the number of new articles tagged to trigger a retraining.