-
Notifications
You must be signed in to change notification settings - Fork 68
Description
We're observing some weird behavior when using REL to match historical placenames.
When we query for 'Zucchabar' and use the following context: "Zucchabar: Stadt in Mauretania Caesariensis", REL returns Miliana as a match with a 91% certainty.
I believe REL uses data from 2019, so I went back in the edit history of both Wikipedia pages to look for an explanation, but couldn't find a clear explanation.
Reading the article Miliana from march 2020 (https://en.wikipedia.org/w/index.php?title=Miliana&oldid=946156179 ) Zucchabar is mentioned seven times, with no link to the actual article on Zuccabar. (there is a subtle spelling difference).
When I check the article on Zuccabar from January 2020 (https://en.wikipedia.org/w/index.php?title=Zuccabar&oldid=937069168) I find seven mentions of Zucchabar, and a few other spelling variants as well as a link to the modern city of Miliana.
Furthermore, the context words occur more frequently in the Zuccabar article than in the Miliana article.
Does anyone have a plausible explanation for this kind of behavior?