Meaning of score

[JueXiuHuang]

Hi,

In Step-by-step extraction tutorial, we can get a concept_instance_pairs object which contains list of concept, eventuality and their score.
What is the actual meaning of the score? Does it means that the probability of eventuality belongs to certain concept?
But tutorial page said that one eventuality can map into several concepts, so I think that the score might represent the weight that eventuality contains certain concept. But the sum of these weight doesn't equal to 1, or just because we didn't normalize the weight?
Which explanation is closer to the real one?

Thanks!!

[seanliu96]

Hi,

Thanks for your great question. The score is calculated at concept.py#L165. There are many ways to get the score(instance, concept). The default is "likelihood", i.e., frequency(instance, concept) / frequency(instance).
More score functions can be found at [1].

[1] Z. Wang, H. Wang, J. Wen, Y. Xiao. An inference approach to basic level of categorization. In CIKM, 2015, pp. 653–662.