Quoting T!M:

Quote:
Quoting Dr Pavlov:

Quote:
I will say, based on my own experience in finding an aberration in less than a week, that it is probably 1 in 100 or even lower. But let's take 1 in 100, that is over 4000 titles

The key is in your first line. You claim that you've found ONE case. One. Better yet, if I remember correctly, you didn't find one at all - maybe you could clear that up by telling us the affected title? Anyway, even if there was one, you now effortlessly manage to extrapolate that to 4000 titles in a bid to scare us, but again: it's based on nothing. Name me ten affected titles (still a drop in the ocean, of course), and I'll shut up. But you can't. Instead, you'd like me to prove a negative, because you know it can't be done...

Yes I found one case and immediately ceased what i was trying to do because i did not want to risk injecting any kind of error factor. One was all I needed to find.

Skip

Quoting Dr Pavlov:

Quote:
I will say, based on my own experience in finding an aberration in less than a week, that it is probably 1 in 100 or even lower. But let's take 1 in 100, that is over 4000 titles, that is far too big an error for my tastes, which I why am refusing all cloned data that has not ACTUALLY been verified against the ACTUAL credits of the particular version. For my own collection that would be over 40 title, or almost 80 if I include my wishlist.. That kind of error is too big. This one reason why I disagree with those who believe good enough is good enough, it doesn't take much of an error in a database this large to create a real mess very quickly, especially if we cloning data without any form of verification other than a cloned profile.

We have also had several of these anomalies documented right here in the Forums.

Skip

I agree with Skip that we need to come up with a system that as much as humanly possible prevents errors being entered in the database.

However, I'm actually very surprised this comment comes from Skip of all people. You are one of the most fervent 'hard data' supporters there are on the forum, to the point that many errors are now present in the database because of:
- spelling errors in credits
- printing errors on covers
- a lack of a uniformed system to transcribe Asian names, both here as well at the distribution studios that release these movies locally (e.g. Yuuji Kishi, Yuji Kishi and Yûji Kishi are three different actors in the database but actually all refer to the same person, just with his name transcribed differently)
- a lack of a uniformed order system in credits with the first and family name of foreign actors

If you look at it, in fact the 'hard data' copy-pasting enters more errors (not errors vis-a-vis the credits but vis-a-vis real life data) then the new system being presented or discussed about here.


Like I wrote a few pages back, I don't mind to use the hard data as the starting point to enter data into the online database. But adding an actor/crew interlinked database, to add the real name and alias of actors and crew will in fact reduce the number of errors in the database, provided submissions require good, recognizable sources.

If anything, I feel the current system is also a form of 'good is good enough': we know there are errors in credits and covers, but since we don't have an infallible system to correct those errors, we take the lesser of two evils and stick to blind copy-pasting of covers and cast/crew. To me, that also seems like a form of 'good is good enough'.

If Invelos can come up with a good system to create a working, linked actor/crew database that can be linked to the credits of the 'hard data', then I'm sure we'll be able to have a database that can not only display a one-on-one copy of the credits, but also an accurate, reliable overview of real actor and crew names, for those that prefer this over hard data cast/crew listings.

Quoting Taro:

Quote:
Quoting Dr Pavlov:

Quote:
I will say, based on my own experience in finding an aberration in less than a week, that it is probably 1 in 100 or even lower. But let's take 1 in 100, that is over 4000 titles, that is far too big an error for my tastes, which I why am refusing all cloned data that has not ACTUALLY been verified against the ACTUAL credits of the particular version. For my own collection that would be over 40 title, or almost 80 if I include my wishlist.. That kind of error is too big. This one reason why I disagree with those who believe good enough is good enough, it doesn't take much of an error in a database this large to create a real mess very quickly, especially if we cloning data without any form of verification other than a cloned profile.

We have also had several of these anomalies documented right here in the Forums.

Skip

I agree with Skip that we need to come up with a system that as much as humanly possible prevents errors being entered in the database.

However, I'm actually very surprised this comment comes from Skip of all people. You are one of the most fervent 'hard data' supporters there are on the forum, to the point that many errors are now present in the database because of:
- spelling errors in credits
- printing errors on covers
- a lack of a uniformed system to transcribe Asian names, both here as well at the distribution studios that release these movies locally (e.g. Yuuji Kishi, Yuji Kishi and Yûji Kishi are three different actors in the database but actually all refer to the same person, just with his name transcribed differently)
- a lack of a uniformed order system in credits with the first and family name of foreign actors

If you look at it, in fact the 'hard data' copy-pasting enters more errors (not errors vis-a-vis the credits but vis-a-vis real life data) then the new system being presented or discussed about here.


Like I wrote a few pages back, I don't mind to use the hard data as the starting point to enter data into the online database. But adding an actor/crew interlinked database, to add the real name and alias of actors and crew will in fact reduce the number of errors in the database, provided submissions require good, recognizable sources.

If anything, I feel the current system is also a form of 'good is good enough': we know there are errors in credits and covers, but since we don't have an infallible system to correct those errors, we take the lesser of two evils and stick to blind copy-pasting of covers and cast/crew. To me, that also seems like a form of 'good is good enough'.

If Invelos can come up with a good system to create a working, linked actor/crew database that can be linked to the credits of the 'hard data', then I'm sure we'll be able to have a database that can not only display a one-on-one copy of the credits, but also an accurate, reliable overview of real actor and crew names, for those that prefer this over hard data cast/crew listings.

ERRORS based on hard data, not uer-interpreted data. What you THINK is an error is for you to deal with locally.

WSIWYT, Taro

Skip

There is way to do this. We should have two Cast and Crew databases.

Primary one would be binded to a movie. Lets say that the key field would be a combination of Original Title, Director and production year for an example. This would be our primary source for cast & crew and shared with all localities and releases.

Secondary Cast & crew would would be binded to UPC and only contain modifications compared to primary cast & crew db and used only if your release doesn't match with primary db data. There you could add localized actors, or if there are any differences compared to primary cast & crew.

These two fields should only be seen in edit mode. When you switch to view mode it would show only one set of cast & crew which would be a combination of those two db's. This would raise the overall db accuracy to whole new level. There are only few cases where you would actually need the secondary db.

Quoting Dr Pavlov:

Quote:
OK, I will take exception to one of your comments. You claim that a Movie is Unique, the BEST that can be said is that a DVD is Unique. We have already proven that Movies themselves are not necessarily unique. Release A may not in many cases contain precisely the same data as release B. How you define release is actually irelevant to this, it could be Release A in Region 1 US versus Release B in Region 2 France or Releases A  versus Release B SE.

This why cloning of data ortrying to set up a single profile for a title is so dangerous because they are not unique.

Skip

If you look at what I wrote:

A Movie is Unique
A MOVIE can have one or many titles
A Person is Unique
A Person can have one or many Credited or Uncredited (supported by documentation) Roles (Cast or Crew) in a Movie Title

I said that a Movie can have one or many Titles.

The Movie identifier can be an internally generated number. I am betting that if the Movie Title matches, then the credits probably match as well. If they don't then a view per UPC/EAN or Disc id can be created to filter out the differences.

As to Cast and Crew, we would only need to link the variants. For example:

John Smith
John R. Smith
John Rchard Smith

No links.

John Smith Born 01/01/1901 in Albania Died 01/01/2000
John R. Smith Born 01/01/1901 in Albania Died 01/01/2000
John Rchard Smith Born 01/01/1901 in Albania Died 01/01/2000

Automatically create a link and consolidate.

Quoting Dr. Killpatient:

Quote:
Quoting Dr Pavlov:

Quote:
NASA lost a multi-million dollar spacecraft because of that kind of lackadaisical attitude. ANY error which can be avoided, should be because errors multiply.

Um, scope?  Nobody is going to die if someone is credited wrong from copying a profile.

Yeah, I was kinda floored by this one as well.  It would probably hold more weight if I wasn't still stumbling across errors almost every time I purchase and download a new profile.  I mean, the system we have now is working so well, wouldn't want to jeopardize that well oiled machine. 

Quoting mdnitoil:

Quote:
Quoting Dr. Killpatient:

Quote:
Quoting Dr Pavlov:

Quote:
NASA lost a multi-million dollar spacecraft because of that kind of lackadaisical attitude. ANY error which can be avoided, should be because errors multiply.

Um, scope?  Nobody is going to die if someone is credited wrong from copying a profile.

Yeah, I was kinda floored by this one as well.  It would probably hold more weight if I wasn't still stumbling across errors almost every time I purchase and download a new profile.  I mean, the system we have now is working so well, wouldn't want to jeopardize that well oiled machine. 

You aren't, you're not looking then. I find errors all the time, sometimes I fix them (depending on my mood, usually I just ignore them. I just found two rather significant errors in Australia-BD, points for anyone who finds them. One should slap you in the face, the other is a bit more subtle.

Skip