Discussions on psych0fred

I think that I understand where you're coming from. What we're trying to do on Freebase is to have all of the data in one place, and only have one instance of each thing. You can and should overwrite the data that we imported from Wikipedia if it isn't clean or correct. I'm just asking that you overwrite the existing Wikipedia data instead of creating new, alternate data. That way, other users don't get confused by multiple records for the same episode, and we're able to tell that the X-Files episode info that we got from Wikipedia has been fixed.

Hi Jake, welcome to Freebase. We consider each X-Files episode that ever aired to be single, discrete concepts. While different sources (Wikipedia, IMDb, etc.) will contain their own entries to represent each of these episodes, they are all attempting to represent the *same thing*. For practical reasons Freebase was, and continues to be, populated with almost every topic in Wikipedia -- this gives us a great starting ground to work from. As it happens, a lot of X-Files episodes were already represented in Wikipedia, which means they were automatically given entries in Freebase too. Additionally, we use Wikipedia extensively to actually populate Freebase topics with as much structured data as we can reasonably extract -- for the most part it's a great and relatively accurate resource. It's important to understand, though, that we really have no special respect whatsoever for the data contained in any one of those Wikipedia articles. If somebody like you comes along and changes data that originally came from Wikipedia -- great. That means we're probably *more correct* than Wikipedia was. Unless the newly contributed data turns out to be wrong, there is no reason to go back and replace it with the old data from Wikipedia. We want to collect the best, most correct data we can find. However, even if a Wikipedia article did contain data that was simply wrong, we *still* want to retain the link to that Wikipedia article. What we're retaining is simply an assertion that "this topic in Freebase has a corresponding entry in Wikipedia". This is useful information to keep because: A - it prevents us from importing the old article from Wikipedia again in the future, and B - it allows for reliable, direct comparison and synchronization between the two sources, which can be useful both for us and for Wikipedia. For this reason, using existing topics that represent the thing you're contributing to, no matter where its data originally came from, is always preferable to creating new entries, and it is never wrong to correct data that is false. Anywhere there are two different Freebase topics representing the same thing, we consider there to be an error. The solution is to "merge" them together, rather than deleting one or the other. In short, we absolutely *prefer* what you added about the X-Files than what Wikipedia would have given us. We just don't want to lose the fact that you and Wikipedia are talking about the same thing. Hope that clears things up a bit.

I'm really sorry to hear that. Freebase is not what I thought it was. In my mind you've simply branched Wikipedia and structured the data in your branch. That doesn't have a lot of value to me, especially since there are wiki APIs and the interface for adding data to a wiki is much more efficient. It would be simpler for me to set up a wiki and structure the data in it as pages for query using an API. The integrity of the data I add is very important, and accuracy is as subjective as your criteria for accuracy. My criteria for the episode guide information I was adding is the opinion of it's creator, which is not apparent to most. For example inferring what a character thinks or in an episode description is the opinion of the person doing the inferring, not the writer of the episode (or else it would be implied or explicitly stated). It may seem trivial to most, but these kinds of leaps are where innaccuracies in the representation of facts come from. If news organizations were this careful there'd be a lot less spin, bias, and political games and more truth in the world.

There is one important distinction between Freebase and a wiki. Each piece of information is attributed as a separate quantum; if you look at a page history, you can see that. With a wiki, one can certainly look at the diffs to figure out when a particular datum was added and by whom, but it’s non-trivial to figure out the attribution for it. With Freebase, not only can we tell you which user added any given factoid, we can also globally revert edits made by malicious or wildly unreliable users once we notice them. This information is even exposed in the API, so a developer could even build a view of Freebase entirely free of assertions from a particular source.

That's an interesting approach, so like every good versioning system you can check out the version of the build you want. Some wikis do work like this too, I've seen some pretty crazy things done with Jotspot by talented developers who know how to make it jump through hoops.

I agree with you that the data in Wikipedia is low quality. We've imported a lot of that sloppy data, and tried to fix it up. You've done a great job of cleaning up the X-Files data. I'm not asking that you favor Wikipedia data over your data. Much of the Wikipedia data is wrong, and it needs to be fixed. What I'm asking is that you change the data on the existing Wikipedia-based records in Freebase instead of creating brand-new records. The Wikipedia-based records can be linked back to Wikipedia, even after you fix or replace the article and data, and this lets Freebase make this data available back to Wikipedia. Creating brand-new records means that it will be difficult for Wikipedia to ever pull that data back in. We really appreciate what you're doing. Keep up the good work! Thanks Colin

I'd prefer if you marked them for merge - there is metadata that connects those articles to Wikipedia that I'd like to preserve. They will get recreated when we next do an import from Wikipedia otherwise.

Hi, the X-Files topics loaded in from Wikipedia have articles that are useful. See the article on the Season 1 episode "Beyond the Sea" topic, for example: http://www.freebase.com/view/?id=%239202a8c04000641f8000000000dbbbd6 It makes sense to upload new data that doesn't exist in Freebase, but when a topic already exists, it's best not to introduce duplicates but rather, add the types and properties that would categorize the topic correctly and thus help users find it in the future. For example, the topic indicated above is already typed TV Episode as part of the X Files show, but without the Season and other properties filled in. Perhaps that's why you didn't know it already existed before your data load.