read some papers, but still do not understand.
remote supervision: by aligning the knowledge base with unstructured text to automatically build a large number of training data, reduce the dependence of the model on manual tagging data, and enhance the cross-domain adaptability of the model.
so the process of remote monitoring is just the process of automating the construction of datasets? Does not involve the training and prediction of the model?
in addition, it seems that in remote monitoring, the concept of bag is put forward, and whenever a relation pair appears in bag, it is marked as a positive example; otherwise, what is the purpose of marking it as a negative example?

ask for advice from our predecessors. Thank you.