How to interpret test_results? about opennre HOT 3 CLOSED

Hi,

model_name_x and model_name_y are the coordinates of the points on the precision-recall curve of the model. They are calculated in the function test in framework.py. More specifically, first we calculate the classification probabilities of all triples (entity_head, entity_tail, relation_type), then sort the triples by their probabilities so we can get many pairs of (precision, recall).

About your last question, I assume you are talking about how we get the dataset. The NYT dataset is a distantly supervised dataset. The entity pairs and the relations came from Freebase, and then are attached to the sentences in the NYT corpus. You can refer to this paper for details.

from opennre.

I understand. How do I understand which entities and relationships are being assigned to a test sentence? Is their a way to label the output?

from opennre.

There is a basic assumption: If two entities participate in a relation (in a knowledge base like freebase), all sentences that mention these two entities express that relation. But it will bring lots of noisy data. You can refer to this paper for details about how NYT dataset was generated.

from opennre.