Kluge:

Let me try and explain using your Wisher.

There are several main reasons why the CLT system does not function properly.
1) It is completely dependent on user data entry. For example in wisher there is a lot of "crossover" data, showing Wisher in The Abyss AND Wisher, Jr. in The Abyss. Now these could BOTH actually be correct, BUT they are alsobetter than even money to be in at least some instances incorrect, and tjhose need to verified by users who own them.

2) Users refusing to provide CLT results is not helpful

3) This one can go two ways and it's hard to tell precuisely. But we have data which is more than likely incorrectly entered. This can happen by users simply ignoring the Rules and doing whatever they please and/or the mass import of old Intervocative data, some Profiles of which have not been touched for YEARS.

I am glad to see that you understand that linking is by and large a local issue. Thuis is exactly why i believ that a Simple association system works best. Though it does have some issues to resolve. With the SA method there is NO priority named that is determined by anything or anyone, all name variants, once they are linked would yield exactly the same list of movies, these could even be prioritized by the user for example the Searched on name could appera first in the list with other variants following based on frequency. There are as i noted some things that would need to be sorted out, but that is true of most anything in software. But, the system as it is currently designed is simply doomed to fail, for all of the reasons i have listed. The only reason i can see to continue the existing system is if ken has something planned regarding it in the future, but right now it is  a mess with a truly horrendous workload on all of us to get it anywhere close to right.

There are, for example, approximately 60 different copies of Ther Abyss and it is conceivable that some of these could indeed list William Wisher, Jr. in the credits. In this case i have only one copy in my collection and that single copy lists William wisher, but the rest of them need to be looked at by the owners and verified. It's a mess. The only way we could have any hope, is that SAME Disc IIDs, would have the same dataset. So, if for example Region 2 UK has the same Disc ID as Region 2 France, Finland and norway, all of that data will be the same relative to everything even Features and languages. But if region 2 Italy has a different disc ID then all bets are off and that one has to be independently verified.

It's ugly, i grant you.

Skip

Quoting Prof. Kingsfield:

Quote:
Kluge:

Let me try and explain using your Wisher.

There are several main reasons why the CLT system does not function properly.
1) It is completely dependent on user data entry. For example in wisher there is a lot of "crossover" data, showing Wisher in The Abyss AND Wisher, Jr. in The Abyss. Now these could BOTH actually be correct, BUT they are alsobetter than even money to be in at least some instances incorrect, and tjhose need to verified by users who own them.

2) Users refusing to provide CLT results is not helpful

3) This one can go two ways and it's hard to tell precuisely. But we have data which is more than likely incorrectly entered. This can happen by users simply ignoring the Rules and doing whatever they please and/or the mass import of old Intervocative data, some Profiles of which have not been touched for YEARS.

I am glad to see that you understand that linking is by and large a local issue. Thuis is exactly why i believ that a Simple association system works best. Though it does have some issues to resolve. With the SA method there is NO priority named that is determined by anything or anyone, all name variants, once they are linked would yield exactly the same list of movies, these could even be prioritized by the user for example the Searched on name could appera first in the list with other variants following based on frequency. There are as i noted some things that would need to be sorted out, but that is true of most anything in software. But, the system as it is currently designed is simply doomed to fail, for all of the reasons i have listed. The only reason i can see to continue the existing system is if ken has something planned regarding it in the future, but right now it is  a mess with a truly horrendous workload on all of us to get it anywhere close to right.

There are, for example, approximately 60 different copies of Ther Abyss and it is conceivable that some of these could indeed list William Wisher, Jr. in the credits. In this case i have only one copy in my collection and that single copy lists William wisher, but the rest of them need to be looked at by the owners and verified. It's a mess. The only way we could have any hope, is that SAME Disc IIDs, would have the same dataset. So, if for example Region 2 UK has the same Disc ID as Region 2 France, Finland and norway, all of that data will be the same relative to everything even Features and languages. But if region 2 Italy has a different disc ID then all bets are off and that one has to be independently verified.

It's ugly, i grant you.

Skip

Skip,
thank you for your kind explanation, but, belive me, sometime I can hardly follow your argumentations.
You are making reasonable assumptions (there are always exceptions) about DVDs with the same Disc ID, however you refuse to make the same assumption about the IMDB mined data in CLT.
In my experience there are very rare instances of credit changes in dvd like "The Abyss", and mostly(but not ever) you can find changes when the distributor translate the credits, and this is not very usual in the last years, with the exeption of asian films.
The chance to make a bad assumptions in this cases is definitively very low, and it's ever verifiable and rectifiable with our peer review system.
Furthermore, in the CLT results can be badly influenced by title and production year errors (and these data can be all verified without the original DVD).

Last but not least: we have rules, when our interpretations of the rules differs, it's up to the reviewers, Ken and Gerry, and I think that we must accept their decisions (or not-decisions), and is perfectly useless to go on with n pages of chats, maybe could be better to do our best for the improvement of our database, but that is only my opinion.

I hope I didn't offend anyone.

I step away for 24 hours and a 11 page thread shows up?  How the [bleep] does that happen?  It's like a new record!

Okay, let me read the last couple pages....

Quoting Prof. Kingsfield:

Quote:
Tell me how remarks like that are supposed to change my mind, Rho. It actually does exactly the opposite.

Skip

Why would I want to change your mind? I have just stated my opinion about team playing.

BTW you have brought up the topic of the larger team in this thread.

Quoting Prof. Kingsfield:

Quote:
Quoting T!M:

Quote:
Quoting Prof. Kingsfield:

Quote:
This one can go two ways and it's hard to tell precuisely.

This one goes only one way: William Wisher.

Quote:
right now it is  a mess with a truly horrendous workload on all of us to get it anywhere close to right.

Thanks to the huge batch of IMDb-mined cast and crew in the database, and some other problems (incorrectly formatted titles, incorrect production years, missing original titles, TV-credits being counted dozens of times due to differing/translated disc and season indicators listed as part of the title fields), the CLT numbers are indeed somewhat of a mess, and that indeed causes what you might call "a truly horrendous workload" for those that try to improve things. I, for one, am doing the best I can.

Tim:

Please., as long as you continue to refuse to document in your notes. stay out of it.. That's trying to be polite

Dr, Professor:

Please., as long as you continue to require more than Invelos does stay out of it.. That's trying to be polite

Then you need to get Ken to change his requirements.... still 

Personally, I am all for proper notes for name commonality. 

I have already, in the last couple of days, done research as a voter on 2 contributions (And I am sure that there are more).  Luckily they pulled their contributions after I PM'd them with my concerns.

With the way that it is set up now, some people are contributing based strictly on what is in CLT, or by results of name variants plug in (for the ones that are still using it). 

While these are tools to help in decision of what name is common, care must be taken in interpreting the data.  These tools do nothing to help prove that A=B, and must be researched outside of Invelos to make these connections.  For example, (not actual credit) how many different John Smith are in the DB.  CLT or namevariants will not show the difference between a  sound designer in the 1990's as opposed to the actor in the 1960's.  CLT and namevariants will not help decide this. 

So to just say John Smith is 480/1190 and John Doe Smith is 120/456 does not necessarily prove the John Smith and John Doe Smith are the same person in any or all credits under each one.  While it may be true that some John Doe Smith is credited as John Smith, we cannot know until a full research and auditing is done.

If we are not going to submit documentation and or proof, via contribution notes or a pinned forum for such information, then we must be sure to do the research and not expect the research to be done by the voters.  It is the responsibility of the contributor to do the research (and in my view make proof somewhere).  After all, the contributor is the one requesting the change to the DB.  Why should the voters be responsible for this.  It should be more like science.  A scientists does his research and  then submits all research for peer review. Then it either accepted or rejected.

On the other hand, many voters are just rubber stamping contributions.  While this may be favorable to the contributor, it does not say anything good or bad about the contribution.  If you are voting on a contribution, you are stating that the contribution is better than the previous entry.  You are basically giving your approval for the new entry.  While, sometimes trust may be earned over time from a contributor, the contributor must be checked to keep his trust.

The arguments that are presented, are unfortunately a result of ambiguity of the rules, and a general mistrust between some users.

The rules need to be clear.

Contributors and voters must perform their respective roles properly

and the screeners need to be proper arbiters.

Sorry for the rant


Charlie