Joshua Taylor presents Discovering Simple Mappings Between Relational Database Schemas and Ontologies Jesse Weaver 2

Question is for the Presentation: Joshua Taylor presents Discovering Simple Mappings Between Relational Database Schemas and Ontologies
Question is asked by: Jesse Weaver
The Question is: Figure 5 shows Marson achieving an average F1-Measure of less than 0.8 in four out of the five cases shown. How good is an average F1-Measure of 0.8? How good is an average F1-Measure in the range of 0.7 to 0.9?
Question answered by: Joshua Taylor
Answer: I'm not sure how good an F1-Meaure in these particular ranges is. What's even more important, though, is that F1 presumes that precision and recall are equally important. Since this system (and most other systems, I'd suspect) aren't yet good enough to cut the human out of the loop, I think a performance metric should probably be based on whichever humans find more difficult to fix. E.g., if humans are good at spotting incorrect mappings, but not missing ones, a performance metric ought to emphasize recall over prevision. Marson does include a consistency checking phase, but as we well know, a consistent ontology isn't necessary a correct ontology, so perhaps precision is more important than recall.

Question answer	I'm not sure how good an F1-Meaure in thes … I'm not sure how good an F1-Meaure in these particular ranges is. What's even more important, though, is that F1 presumes that precision and recall are equally important. Since this system (and most other systems, I'd suspect) aren't yet good enough to cut the human out of the loop, I think a performance metric should probably be based on whichever humans find more difficult to fix. E.g., if humans are good at spotting incorrect mappings, but not missing ones, a performance metric ought to emphasize recall over prevision. Marson does include a consistency checking phase, but as we well know, a consistent ontology isn't necessary a correct ontology, so perhaps precision is more important than recall. s precision is more important than recall.
Question answered by	Joshua Taylor +
Question asked	Figure 5 shows Marson achieving an average F1-Measure of less than 0.8 in four out of the five cases shown. How good is an average F1-Measure of 0.8? How good is an average F1-Measure in the range of 0.7 to 0.9?
Question asked by	Jesse Weaver +
Question for the Presentation	Joshua Taylor presents Discovering Simple Mappings Between Relational Database Schemas and Ontologies +