[request][denied] recommendations suggestion - Anime-Planet forum
Why register?

make an anime and manga list, and more! all free!

Theme
Old 10-14-2009, 01:11 PM   #1 (permalink)
Narumon Z
 
Join Date: Oct 2009
Posts: 26
Lightbulb [request][denied] recommendations suggestion

I cannot think of a better place to place this so it's going here. Just skip the 1st paragraph if you don't want to read all this. Anyway, this site seems to be dedicated in creating a place and community that will help its users find anime and manga that they would enjoy. (the rather excellent rating and recommendation functions along with the forums and review syestems) I don't know any other site that comes close to this site but i do have a simple suggestion on making the review and recommendation systems work together. Right now you can rate and your rating will be factored into the overall score. that's nothing new and then also there is a recommendation system, where your recommendations are tallied for a specific anime. I think both of these functions can be combined into the search function.

You know how netflix and youtube among others have personal recommendation functions; i think it would be rather easy and useful for one to be implemented on this website. This would make it easier for users to process all the information in order for them to determine which anime they might want to watch next.

Basically all i'm suggesting is creating a personal recommendation button that would recommend anime based on your own personal ratings of anime along with the recommendations of the entire site. so like the personal recommendations on the movie websites.

It should be rather easy to implement and not take any real space up. Basically it would take all the animes you rate and find what people recommended for each anime. The purpose would be to take how strongly you liked an anime and multiple it with how similar it is with other animes. it would take what ever number you rated that anime and use that coefficient to multiple with the recomendations. so a 5 would be 1 and a .5 would be -1.

Quote:
Conversion Chart:
5: 1
4.5: .8
4: .6
3.5: .4
3: .2
2.5: -.2
2: -.4
1.5: -.6
1: -.8
5: -1
In order to find out how similar an anime is with another it just takes the recommendations for that anime and divide it by the total recommendations for that anime. so if 4 out of 10 people thought you should view a certain anime that means it is .4 or 40% similar. Dividing this number makes sure that the most popular animes don't skew the results and it should be extremely easy to count how many recommendations for a particular anime.

It would do this for all the anime you rated and add it together and provide a ranked list. so if you got .4 for one anime and .2 for the same anime but from a different review it would be .6. If you want to actually display the numbers you can multiply by 100 or something to get a percent. You could even add it apart of the extended search function so you can limit your results to just mecha or whatever.

Because this is based on all things all ready present on this website it should take up no space and also be rather easy to implement.
Cyan is offline   Reply With Quote
Old 10-14-2009, 01:13 PM   #2 (permalink)
Narumon Z
 
Join Date: Oct 2009
Posts: 26
Default Re: Site Suggestion

I hope this doesn't seem too confusing. It's all a rather simple idea. it's just hard to communicate how it would work, based on words.
Cyan is offline   Reply With Quote
Old 10-14-2009, 01:45 PM   #3 (permalink)
Moderator
 
Cetonis's Avatar
 
Join Date: Jul 2006
Location: Vernon, CT
Age: 26
Posts: 2,750
Default Re: Site Suggestion

I think this sort of thing has been brought up before; a couple problems off the top of my head:

1) The simple list of recs is only a small part of the recommendations feature here; the real meat is the fact that there are detailed, written out reasons for people to look at. This suggestion would completely bypass that, and be of lower quality as a result.

2) This sort of system would have an extraordinary amount of popularity skew. Shows that are seen less are naturally going to be recommended less. While that is also true for the lists of recs on individual entries, it's easy enough to browse to the later pages, read the reasons and perhaps recognize that a show is only recced once or twice because it isn't exactly FMA. The sort of system you propose would have all such not-so-popular series buried down at lower numbers, where the nature of the feature (the stuff at the top is the stuff you're most apt to like, supposedly) implies that they should be ignored.

All told, people are much more likely to find things they enjoy by browsing the recs as they are - reading the explanations, checking out synopses of things that show up on the later pages that they haven't heard of, etc. The sort of list you mention would be interesting from a curiosity standpoint, but it's not worth the kind of money it would take just to do something the site already does in a better way to begin with.
__________________
Cetonis is offline   Reply With Quote
Old 10-14-2009, 01:51 PM   #4 (permalink)
Cross-eyed Cyclops
 
Drahken's Avatar
 
Join Date: Nov 2007
Location: 1985
Posts: 3,017
Default Re: [request] recommendations suggestion

Sothis has often stated that she's against any form of automated recs, anime-planet is focused on quality, so she wants all recs to be made by real people.
I'll leave this thread in place for the time being, just in case sothis might have something to say (once in a while she does change her mind on things, but only very rarely). I wouldn't get your hopes up though.

Suggestions & feedback are always welcome though, as long as it's nothing idiotic like "GIMME FREE PR0N!". ;)
__________________
*Warning: The above post may be highly opinionated, read it at your own risk.
Wham! Bam! It's the dino-jam!


Halloween Music
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tag Definitions . | . Answers to Common Questions
Drahken is offline   Reply With Quote
Old 10-14-2009, 02:11 PM   #5 (permalink)
Narumon Z
 
Join Date: Oct 2009
Posts: 26
Default Re: Site Suggestion

to Cetonis.
1.
you can just click the actually anime and read the recs from there. it should have recs from multiple anime that you've already saw. so it would be inferior quality unless that user made it so.

2. i suggested you divide it by the total number of recs to avoid that. the unpopular ones are already near the bottom and can be pages deep. so they are being proportionally represented. the popular ones wouldn't be over represented based on the fact it's divided by the whole.

3. Even if my idea was implemented i'd still would look through the recommendations and reviews. a simple equation isn't going to do much for you and take the place of a thought out review. it might tell you to go north but it isn't going to tell you how far north you have to go.
Cyan is offline   Reply With Quote
Old 10-14-2009, 03:01 PM   #6 (permalink)
Overlord
 
sothis's Avatar
 
Join Date: May 2002
Location: Seattle, Washington
Age: 32
Posts: 15,674
Default Re: [request] recommendations suggestion

i just want to say one thing - no one is able to say 'it would be easy' or 'it would be simple' without knowing how a massive system like AP works (meaning, unless they know specifically how AP works). even minor changes (far less minor than what's being asked here, which is actually a massive amount of work) can cost in the hundreds or thousands of dollars (based on hourly rate). implementing a multi part algorithm based on both the personal anime list table, recommendations, anime/manga and others is by no means 'easy' to implement - not to mention the fact that the user anime list table alone has millions and millions of rows alone, thus this sort of algorithm would have to be optimized heavily and carefully to ensure the server doesn't crawl/die.

ideas always seem 'easy' - actual implementation and time/money spent is not.
__________________

Anime-Planet.com - anime | manga | reviews

Adding Anime - Synopsis Templates | Twitter | MySpace

[sixth sense voice] I SEE BRITISH PEOPLE - Control my signature or avatar!
sothis is offline   Reply With Quote
Old 10-14-2009, 03:38 PM   #7 (permalink)
Narumon Z
 
Join Date: Oct 2009
Posts: 26
Default Re: [request] recommendations suggestion

Quote:
Originally Posted by sothis View Post
i just want to say one thing - no one is able to say 'it would be easy' or 'it would be simple' without knowing how a massive system like AP works (meaning, unless they know specifically how AP works). even minor changes (far less minor than what's being asked here, which is actually a massive amount of work) can cost in the hundreds or thousands of dollars (based on hourly rate). implementing a multi part algorithm based on both the personal anime list table, recommendations, anime/manga and others is by no means 'easy' to implement - not to mention the fact that the user anime list table alone has millions and millions of rows alone, thus this sort of algorithm would have to be optimized heavily and carefully to ensure the server doesn't crawl/die.

ideas always seem 'easy' - actual implementation and time/money spent is not.
I know that running a site especially of this size must be extremely difficult. That's why i tried making my idea as basic as possible, i didn't mean to suggest running this site was an easy thing to do. Like you said i'm not aware on how the site is set up so it seemed easy. oh well it was merely a suggestion and based on how difficult any change in the system apparently is, i don't really think my suggestion warrants further thought.
Cyan is offline   Reply With Quote
Reply

Bookmarks

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are On



All times are GMT -7. The time now is 12:32 AM.

Powered by vBulletin® Version 3.8.6
Copyright ©2000 - 2013, Jelsoft Enterprises Ltd.
Search Engine Friendly URLs by vBSEO 3.6.0