Fæ skrev 2013-11-27 15:24 Small error rates are a real challenge. My experience on Commons for large bot work has been long discussions around quality complaints where the level of error was *well below 1%*.
Very interesting you also mention this level of problematic articles. We has found this magnitude in two bot project and our Wikidata project, all because of some problem with source data. We find they can not be found during test rounds, as we do not know what type of problems will turn up (and the articles generated are correct in relation with the source), and it can often take months for the community to spot them (like an erroneous name for a river leaving a lake).
When we entered articles manually 0,5% error level would be more then excellent, but when semiautomated 0,5% can by be seen by the community as problematic (but hardly by the Wikipedia readers).
And to lower the problematic ones semimanually from 0.5 to 0,1 can sometime take as much time as the total generation effort
We at sv:wp still have to come to terms with this, but I expect w at the end must live with this level of problematic articles (not necesary erronous) , and solve them case by case. We are a relative small community (1/40 of en:wp) and have to be pragmatic if we want to give our readers a lot of valuable information
Anders _______________________________________________ Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, <mailto:wikimedia-l-requ...@lists.wikimedia.org?subject=unsubscribe>
