To help us pursue T244192: Newcomer tasks: ORES ontology mapping and score thresholds, @Halfak is going to give us recommended thresholds for each topic model in each language that he expects will give us about 70% precision or better.
He'll also put together some instructions for how we can determine and adjust such thresholds in the future.
@Halfak -- if you get the sense that there are lots of articles that can be captured at 70% precision, then maybe you can recommend thresholds that will get us higher than 70% but with still sufficient recall to supply plenty of articles. Or we can try and adjust from there.