Currently, the rows extracted use an abstract model where each mention can be qualified with a true triple in DBpedia. That is just relevant to get training data. When we are extracting rows to make actual predictions, getting those true triples mess times to get irrelevant information.
Currently, the rows extracted use an abstract model where each mention can be qualified with a true triple in DBpedia. That is just relevant to get training data. When we are extracting rows to make actual predictions, getting those true triples mess times to get irrelevant information.