scientific journals are only increasing and doing synthesis traditionally is extremely difficult and tedious. using ml, it’s possible to identify and classify relevant literature.
identification and classification is similar to how we produce systematic maps. can be used to identify where we have lots of evidence, where there’s evidence gaps, and the maps can provide starting points towards more detailed systematic reviews
main benefits:
using nlp to look for scientific texts is huge, so here are some other things I’d like to try:
unsupervised learning - when theres no specific set of categories into which we want to classify documents - unsupervised learning could be helpful
examples of climate lit:
Systemic reviews → instead of simply screening documents to make predictions → prediction can be used to decide which docs to screen