posted this on May 10, 2011 16:29
Fotopedia relies strongly on Wikipedia to organize its albums. Wikipedia and fotopedia are living objects and even if their consistency is improving continuously, you may have noticed some discrepancies in the data such as:
articles appearing empty in fotopedia
two articles sharing the same pictures, making it difficult to avoid irrelevant flagging from one side or the other
We do our best to fix as efficiently as possible any inconsistency in data we import from wikipedia as well as in fotopedia data. But sometimes it gets tricky, and we fall on issues that involves more than one wikipedia language.
You may have noticed that we are getting closer to make fotopedia speaks several languages. http://fr.fotopedia.com has been discreetly opened a few months ago, allowing to browse the french fotopedia, based on the text from the french wikipedia and sharing the pictures of the english fotopedia. It is not fully translated yet, the user interface is still in english, the search engine is not yet plugged in, but we are working on all this. Several other languages are being integrated too.
But that comes to a price. As we try to share pictures between all these wikis, we need the wikipedia “interwiki” links to be as accurate as possible. These links are visible on the left column in the wikipedia pages and point to articles in other languages dealing with the same subject. Sometimes these links are not as accurate as we would like them to be, and even sometimes completely wrong. In such cases, it is very difficult for us to automatically make sense of them, and we cowardly ignore these articles instead of taking the risk to moving pictures around to the wrong fotopedia articles.
As of today, we are impacted by a few hundreds of instances of these links being half-wrong or just wrong. We think the way to go is to go and fix the links directly in wikipedia whenever it is possible. Fotopedia can then catch-up at the next wikipedia import — we import wikipedia data every time wikipedia export it, that is, every month or so.
We are thinking hard about how to deal with that. We will need some help, as this task requires to actually understand what the articles are speaking about, and we do not have in-house knowledge of all the languages that may have to be dealt with. Even if we will try to provide as much hints as what links have to be fixed, in the end, a human brain will have to decide. Our moderation team is aware of these issues and will try to deal as smoothly as possible with the possible conflicts.