Quoting northbloke:

Quote:

I think Hal's idea has the most merit.

I also answered to Hal that I agreed with his idea. But that needs a complete change of the database, and I'm sure we'll wait for that for years. So I propose a less effective solution, but which could solve a large part of the problem as soon as tomorrow, if Ken just accepts to modify the rule.

Quoting surfeur51:

Quote:
Quoting VirusPil:

Quote:
There's also one thing to remember: there are not just names that distinguish in accent or not.

Think of The Rock for example. What is is his right name? For some users it might be The Rock, for some Dwayne Johnson, or ...

Already answered that in my first post, page 1.

Must have overread this... 

... and now a second time overread it. 

Quoting VirusPil:

Quote:
But we have lazy users. That's why I have thought about an idea were the "lazy" users can take afford of the correctness of some other users. How would you make sure that on every contribution a big search for the correct name has be done? 

I'm afraid I've got to agree with VirusPil again. Before we had Ken's "clarification", I made some submissions and some users simply refused to accept Gerri's ruling and voted no. Even if Ken were to turn around and say we were allowed to use local capitalisation rules, we'd still have people vetoing the data, we'd get no further forward. 

Quoting surfeur51:

Quote:
But I also think that if rules precise that in case of capitalized names, the contributor have the responsability to check correct conversion, most people will try to do their best.

I think if you force people to check each and every name, you'll reduce the number of people contributing. By limiting it to what's present on screen, it's a simple process of duplicating what you see. Besides, it's often easy to find different information on different sites. Even official sites aren't always an option as not all people have them.

I really think that with regards to accented names, there's no fix. What we have is far from perfect, but allowing people who do know the difference between François & Francois to link the two together via Credited As is a way of fixing it to the best of our abilities.

Quoting VirusPil:

Quote:
My idea in my second post of this thread was to take the system we have got and the informations we have got and search for a solution which doesn't need a complete new structure.

The biggest drawback I can see with doing it this way, is we would need to fill the database with BY and PY (production year  ) information. Otherwise all the more common names are going to get merged together and we'll get fake "common names" chosen and people linked who shouldn't be.

Quoting Ace_of_Sevens:

Quote:
The problems we have now are due to the structure being broken. We won't be able to get more than marginal improvement without an overhaul.

Couldn't agree more! 

I don't think we will get any meaningful linking, unless we get away from a thinking about it in names.



We need to think of it more in How we associate.  We really don't care if this is the same name as these other 10 variants.  The only thing we care about is "Are these 2 people the same".

This is a 1 to 1, not 1 to many relationship

I am not a very good diagramer, but here goes (very simplified)


                                          KEY1
      J. Doe --  John Doe -- J. R. Doe --  John R Doe -- John Robert Doe

                                          KEY2
      J. Doe --  John Doe -- J. R. Doe --  John R Doe -- John Randall Doe

Here we have 2 distinct people, represented by Key1 and Key2.  The names below them are WYSIWYT.  What name you enter is irrelevant.  No need for BY, or correct names, or common names,  the only thing that matters is the relationship between he name you entered and the key to associate it with.  In this scenario, common name would automatically be determined by the relationship of the number of entries to the key.

When you enter a name into a profile you are auditing, the system would use a soundex or some other filtering subroutine to determine possibilities, and present them into a list.  So if you type in:

John Doe as Role 1

A list would pop up and give you an option of like

1  J. Doe --  John Doe -- J. R. Doe --  John R Doe -- John Robert Doe
2  J. Doe --  John Doe -- J. R. Doe --  John R Doe -- John Randall Doe

Of course the more info you have, the closer the list becomes.  So typing in John R. Doe, would of course limit the list.  At this point, you would be able to click on a select the appropriate one.  If you did not know the right one, you would be able to look at the credits associated with each one, and then make your selection.  the entry in the DB would be John Doe (key?).  Now you have an association.

It would require work, but when fully implemented, cast and crew would be linked properly.  It would not care if you accented or not.

Now, I do not know how to implement this.  I can picture the structure, but not knowing how kens DB is designed, it may very well be a massive overhaul.

well MHO

Charlie

I agree with you Martian. There are several different approaches possible, but something needs to be done.