All loans on Kiva.org are assigned tags that give lenders more search criteria. Examples include #SingleParent and #WomanOwnedBusiness. In the past, 100% of tags were administered by a volunteer group. This system uses a Random Forest, along with a series of hard coded rules, to automatically apply tags to certain loans. Volunteers can easily remove eroneous tags through a website.
DataPrep.ipynb prepares training data. RandomForestModel.ipynb contains the training code using a Random Forest operating on loan descriptions with a bag of words approach. Model.ipynb was an experimental model using an LSTM, but it did not perform as well.
The Autotag directory contains the production code.