[Anime Hint] - Feature Requests
Moderator: AniDB
There were a couple of reasons not to do this.e-Viper wrote:Seems pogic that the auto-filtering of hentai should be dependant of the flag in the profile. So every person can choose what he/she wants
1) There were explicit requests asking for it this way.
2) It would require a database query to build the search screen. The page has been redesigned to reduce load on the server as much as possible.
3) Just because someone may be interested in Hentai, that is not necessarily what they want to have it listed by default.
4) In general, hentai are usually rated lower than regular animes. This means that they usually do not show up in a mixed listing anyway. I believe that most users will do queries either looking for or filtering out Hentai, few people will have mixed results, so the default may as well be by filtering it out. You have to change the settings anyway to limit the search to Hentai.
That is why Hentai was put at the top of the list, so it would be visible (and obvious) that hentai was selected so that users can deselect it easily if they want.
Then an additional checkbox outside the selection box would been better or an option none underneat the hetai one.egg wrote:That is why Hentai was put at the top of the list, so it would be visible (and obvious) that hentai was selected so that users can deselect it easily if they want.
/me wonders how many users will get the CRTL+click (desides it being annoying to use mouse and keyboard together)
you can't have it both way.egg wrote:3) Just because someone may be interested in Hentai, that is not necessarily what they want to have it listed by default.
But if it's not possible to do DBquery for the profile flag then it doesn't matter anyway
Who's that?Rar wrote:Skywalker.
*hrr hrr hrr*
I really did not understand it when I read it. I'd understand an o/p typo but p instead of l? How'd you manage that? Fat fingers?
But this way or the other - if you read your posting before you hit the "submit" button, you really have to stumble over that typo, it really hits you in the face. I thought it was a word I haven't heard yet and asked leo.org and TUC Dict for it before asking the question.
As i see from page and html-code buttons 'Get hint/Update options' in 'Simple:' and 'Advanced:' parts are the same.
I would remove misleading 'Simple:' header (it works the same as advanced, which is not so simple now) OR made button in Simple part worked different (maybe setting options to default and invoke search).
I would remove misleading 'Simple:' header (it works the same as advanced, which is not so simple now) OR made button in Simple part worked different (maybe setting options to default and invoke search).
That was put in by exp so that users would not get overwhelmed by all the fields and they could just hit the button for the results. What I am working on now is having the advanced fields hidden, and if then they can be expanded by the user. You can see a mock up here, note the hint and links do not work, just the hide/show links.Elias wrote:As i see from page and html-code buttons 'Get hint/Update options' in 'Simple:' and 'Advanced:' parts are the same.
I would remove misleading 'Simple:' header (it works the same as advanced, which is not so simple now) OR made button in Simple part worked different (maybe setting options to default and invoke search).
[Edit: If someone with Netscape 6 (or earlier) can test this, I would appreciate it. Otherwise this won't be supported for older browsers. I have tested it with IE, Maxthon, FireFox, Opera, Mozilla and Netscape 7.]
OK, the new interface has been implemented, please let me know if there are any issues.
I have a question for people, in the logic for finding similar users I am going to implement Pearson product-moment correlation coefficient as a more statistically sound approach to finding similar users.
This has the advantage of finding similar patterns in the voting and not depending on the values themselves. For instance, if I vote 10, 9, 8, 7, 8, 9, 10 on group of animes and another user votes 7, 6, 5, 4, 5, 6, 7. Under the current system the user would get a zero value, but with Pearson, it shows a very good match because the pattern is the same. Pearson can also resolve the opposite case where the numbers are close together but the votes show a conflicting pattern.
One thing is the formula falls apart when comparing between two users and all of the votes of one of the users are the same. For instance, although I have voted on 73 anime, I have voted 9 on six of those. Let's say that another user has only those six anime in common with me, if I run the formula I end up with a divide by zero.
There are two ways of trying to handle this:
1) I can completely ignore these users since having flat votes does not have enough information to build a good pattern.
2) The other way is to fall back on the old formulas in these cases, so if the user has votes close to 9, then they would have a high score, but if the votes are farther off, then the score would be lower.
Are there any suggestions?
I have a question for people, in the logic for finding similar users I am going to implement Pearson product-moment correlation coefficient as a more statistically sound approach to finding similar users.
This has the advantage of finding similar patterns in the voting and not depending on the values themselves. For instance, if I vote 10, 9, 8, 7, 8, 9, 10 on group of animes and another user votes 7, 6, 5, 4, 5, 6, 7. Under the current system the user would get a zero value, but with Pearson, it shows a very good match because the pattern is the same. Pearson can also resolve the opposite case where the numbers are close together but the votes show a conflicting pattern.
One thing is the formula falls apart when comparing between two users and all of the votes of one of the users are the same. For instance, although I have voted on 73 anime, I have voted 9 on six of those. Let's say that another user has only those six anime in common with me, if I run the formula I end up with a divide by zero.
There are two ways of trying to handle this:
1) I can completely ignore these users since having flat votes does not have enough information to build a good pattern.
2) The other way is to fall back on the old formulas in these cases, so if the user has votes close to 9, then they would have a high score, but if the votes are farther off, then the score would be lower.
Are there any suggestions?
If 'min.vote' is set to high value, situation like this does not mean another user voted only 6 times 9 (he can have also more different votes, but not counted)?
Then if there is another user wchich has 5x9 and 1x8 from those animes, the 'first' another should be treated as more similar.
To avoid division by zero i would add one 'dummy' vote with 1 (or 9) difference, so this another user would have weight the same as if he had 6 votes of 9 (like you) and 1 vote different with 1 from yours. Or give him weight the same, as if he had 5 votes of 9 and one of 8.
Then if there is another user wchich has 5x9 and 1x8 from those animes, the 'first' another should be treated as more similar.
To avoid division by zero i would add one 'dummy' vote with 1 (or 9) difference, so this another user would have weight the same as if he had 6 votes of 9 (like you) and 1 vote different with 1 from yours. Or give him weight the same, as if he had 5 votes of 9 and one of 8.
The situation that I was describing happens regardless of the other settings. Each user could have voted for a number of anime with wide variety of votes, but the animes that they have in common just happen to have all of the same vote by a particular user.Elias wrote:If 'min.vote' is set to high value, situation like this does not mean another user voted only 6 times 9 (he can have also more different votes, but not counted)?
That is one way of handling it. This is an interesting concept, I will have to play with it to see how it works out.Elias wrote:Then if there is another user wchich has 5x9 and 1x8 from those animes, the 'first' another should be treated as more similar.
To avoid division by zero i would add one 'dummy' vote with 1 (or 9) difference, so this another user would have weight the same as if he had 6 votes of 9 (like you) and 1 vote different with 1 from yours. Or give him weight the same, as if he had 5 votes of 9 and one of 8.