I would like to also go with film based credits rather than DVD based credits (so we don't have to duplicate effort for all the different localities and releases).  I know there are few exceptions where credits differ based on locality, but benefits would outweigh the downsides.  If there is a simple way to make exceptions, even better.

Quoting T!M:

Quote:
I wouldn't presume to know how to handle it from a programmer's point of view (a unique ID for everyone certainly seems a logical step), but what I'd like as a user is, ideally, that when I'm auditing a DVD, and I enter a name - say "John Doe", is that the program asks me which "John Doe" I mean. I'd like the program to show me a list of the existing people named "John Doe", including people whose real or common name may not be John Doe but who have been credited as such. If at all possible, it would be great if something extra was thrown in to further identify the various John Doe's, like maybe including two or three of their film titles, their production years, or when it's about the crew section, maybe the job category in which they're mostly credited. Basically, when I type "John Doe", I'd like to be asked whether I meant the actor that acted in this-and-that, the sound mixer who worked on so-and-so, or the editor that worked on such-and-such. I'm afraid I'm asking a lot here... 

I DO presume to know how to handle it from a programmer's point of view. But to be fair, I think the way to approach this is to figure out how to make it as simple as possible for the end user.  If you start from a programmer's point of view, you'll end up with something that works fantastic that no one can figure out.

With that said, I think T!M is dead on.  I noticed the other day my Mad Men profiles have Elizbeta Moss.  Not sure if that's a mistake to be corrected yet or an alternate spelling.  But it's a good example.  From an end user point of view, I want to enter something like Elizabeth Moss and have the program display all current Elizabeth Moss' in the database.  It should also display any alternate spellings / existing links if they exist.  I'd like to see: Elizabeth Moss (AKA Elisabeth Moss, Lisa Moss, Beth Moss, etc).  I would use code from your CLT tool to order the names from most popular variations to least popular.

From there you could take it a step further and use a Sounds Like algorithm to find similar names. They're fairly simple to implement.

It should give me more than a name.  I should be able to see what that person has worked on previously.  Ideally it would list a couple of their biggest titles with the option to see all titles.  What they're most known for.  But that would require probably finding which titles are the "most owned" by people.  I think that would have to be automated or we would never agree what someone is most known for. This could be what they're most know for overall with the ability to drill down to see every title you own with that actor.  If we're retaining the existing birthyears, those should be displayed as well for identification purposes.  I think that there would handle a huge chunk of problems.  At least for big names.

I'd also display their photo if one is available.

Initially I wouldn't place too many restrictions around linking or unlinking people.  Unless it proves to be a problem.  Someone should be able to say that John Doe from movie A isn't the same John Doe from movie B.  The voters ought to be able to catch most mistakes like that.

That's all I can think of for the end user.  While it's easy to let my imagination run wild, I think it should be a simple interface.  If it becomes too convoluted, you'll end up with frustrated users.

From the technical side, just some thoughts off the top of my head.  This is just a first stab.  I think I would have an Actor table with 2 IDs.  The first would be a Person ID, the second would be a Variant ID.  So with my previous example, Elisabeth Moss might be Person ID #1.  Elizabeth Moss would also be Person ID #1.  Each spelling would get a different Variant ID.  Other fields would include the names.  And if we retain birthyears, those would be there too.  Perhaps a pre-computed "sounds like" value should you decide to implement that for faster searching.

Also in an ideal world, if Elizabeth Moss is a typo, when it's changed, Eliszbeth Moss would be removed from the table if that name is no longer connected to any existing profiles. That would help a little to keep the database clean and keep the potential variations smaller when looking someone up.

So if I enter Elisabeth Moss, it would select any exact matches on that name.  You would also pick up all the variants on the spelling.  If I pick one, that person ID and variant ID get saved in the movie profile.  If I determine it is a new person with the exact same name, they get a new person ID and variant ID.  If you determine that Elizabeth Moss is not actually Elisabeth Moss, then you change the existing Person ID and Variant ID to the proper person or make them a new person.

I'm be happy to ponder the technical side further and consult with some DBA friends, but I suspect you really don't need such assistance.

Naming of headshots should reflect linking (use of ID ?). It could be something like "givenname_surname_ID"
In fact ID field would replace in database BY field, with of course different functions.

At present time, if a user has Zhang Ziyi in his local, and another user has Ziyi Zhang, we get two headshots when we share headshots databases. It would be great to have only one. So we can compare Zhang_Ziyi_56789 to Ziyi_Zhang_56789

If you don't want to be downloading the entire cast/crew list to your local database, then the linking of new names will have to be done by the contribution system.

Say you have John Doe #1 and John Doe #2. You profile a new movie, which you know has John Doe #1 in it, so you select that one from your list when entering. Now, also there is Jane Doe, and you have a Jane Doe #1 but not the correct one. You can enter Jane Doe as a new variant and program gives it a local unique id. When contributing, John Doe #1 is automatically accepted, but for Jane Doe, you are given an option to choose which of the online database's Jane Does are correct, or the option to create a new one.

Obviously the contribution system is going to have to give you list of titles that each Jane Doe is profiled in to make it easier to select the correct one.

I don't think it will sensible to completely remove the CLT, but you can at least change it to give the online variant list

This is what I posted HERE a while back along with a poll and discussion:

Here's how I see it possibly working:

Eliminate the "As Credited" function of the program as well as the CLT.  All credits would be entered WYSIWYT.

Ken would create a "Linking" table which would contain all variations of a person's name.  This table could be initially populated by Ken by extracting the current "As Credited" data from the main database.  This means that all the work done over the past couple of years trying to link names using "As Credited" would be preserved and we would be starting this system with a significant number of names already linked.  This table would become part of the download files whenever you "Update" your local database.

Ken would have to modify the code for the local application such that when you double-click on a name in the cast or crew section, the program would find the exact name that you clicked on in the table (referenced above) and list not just the profiles with an exact match, but profiles with matches to all name variants listed in the table for that name.

The program would link all instances within your local database automatically without making you manually go to each profile and update the credit to include "As Credited" data.

For new links, users would submit "Link Requests" through a separate contribution process.  These requests would require sufficient supporting evidence that the two linked names are in fact the same person.  These "contributions" would be voted on the same way as profile contributions today, and ultimately reviewed by the screeners and approved/rejected as appropriate.

We all know that the current main database is riddled with IMDb data.  A simple linking system wouldn't care if the actual credits in the main database are "As Credited" or not.  It would only need to know that credit A equals credit B equals credit C, therefore, a "cleanup" of the main database is not required to make this system work immediately.

In all fairness, I need to point out one downside of this system.  Take for instance "Al Smithee".  Today, with as credited, "Al Smithee" has been identified to be an alias for more than one person (which is in fact the case) and only the profiles that are associated with the same person will get linked.  In the simple linking system, this cannot be done, and all people associated with the name of "Al Smithee" will get linked together.

Personally, I believe this scenario is so infrequent, that the benefits of a simple linking system far outweigh this one small flaw.  And knowing this community, it is entirely possible that someone here can come up with a solution to that problem, too.

To really put a cap on this, if we collapse the name fields to a single name field, we'd have a near perfect system that would eliminate 75% plus of the arguments in the forum. (I think that's a good thing????).

Please feel free to shoot this full of holes!!!

Fantastic that something like this is finally happening!    And everyone's brought up some nice ideas already!

Quoting surfeur51:

Quote:
Quoting xradman:

Quote:

I don't know about Flipper, but Madonna and Cher are stage names IMO and belong in the Given Name field.

For sorting reasons, those stage names should be better in surname place.

This doesn't really have anything to do with the thread persay (and I'm not trying to derail the thread) but we have to be careful when talking about things like this. Madonna is NOT a stage name. It's her given name. Ciccone (which she's been credited as only a single time in her career) is her surname. Gotta make sure when we get a stage name, it's actually a stage name. I do however understand the need for proper sorting, so it could be as simple as if there's no surname entered, it simply gets sorted by the only entry. Putting a first name into the surname field kind of defeats the purpose (in those cases only) of going to only two fields to properly enter given names into a given name field and surnames into a surname field.

So I'd like to see: A single name = being sorted by that name without having to enter it into the incorrect field.

Quoting Merrik:

Quote:

So I'd like to see: A single name = being sorted by that name without having to enter it into the incorrect field.

I agree. In fact, when I spoke of stage names, I meant unique names or group names. John Wayne should be treated as given name/surname. Unique names may be the given name of the person, or his (her) surname, or a specific stage name (Miou-Miou, Bourvil, Fernandel...). We have also to treat group names (The Beatles). My concern is to find Fernandel among the F, not before the A.

Some ideas by different users have been posted here.

I'll post my current idea this evening.