Dennis Ouwendijk (Customer) asked a question.
In our model we have a list of basin names and in the current logic we're only looking at the beginning of the document for the classification. We'd like to enhance that logic to also look at the rest of the document but only classify it with the most mentioned basin names, as often references are included that are irrelevant for the classification. It should be relative to the total number/count of basin names mentioned throughout the entire document, see the example below for the expected outcome:
Total basin name mentions throughout the document: 5
3x Permian Basin (expected score of 0.6)
1x Sydney Basin (expected score of 0.2)
1x North Sea Basin (expected score of 0.2)
Finally, we'd like to use a cutoff of 0.5 to only display the top mentioned basins.
What would be the best approach to this?
Hi Dennis,
Sorry, about the late reply.
I think you can do something like you wish but putting a count on a foreach rule such that the rule only gets to that score if it meets the number of matches in the bit of the document you are looking at. Obviously, you could also try to restrict which mentions you are firing in the first place, by filtering them out of they do not have any supporting evidence (or negative evidence) surrounding them.
See
https://portal.smartlogic.com/docs/5.6/classification_server_-_rulebase_reference_guide/attributes/foreach
which says
"If the value of foreach is set to 1 then the score will increase from the defined weight, whereas if foreach is set to be greater than 1 the weight defined on the rule will be reached when it has found the number of instances set on the foreach attribute."
You can also have a cutoff on the category rules at the top of the rule tree in your basin template - by specifying the cutoff to be 0.5.
See
https://portal.smartlogic.com/docs/5.6/classification_server_-_rulebase_reference_guide/attributes/cutoff
I think you would need to speak with professional services, though, to get a better idea of this would work.
Cheers!
Akan