[removed151214]

Only when the user specifically enables it - this is NOT something I would ever have enabled by default - and only until I get time to look better into TMDb's api to do it better.

[phil65]

Only when the user specifically enables it - this is NOT something I would ever have enabled by default - and only until I get time to look better into TMDb's api to do it better.

The TMDB API doesnt allow to do it any better, that´s why I was wondering I feel that is over the top considering the needed amount of calls / bandwidth esp when you´re weighing them lowly. What would make more sense to me would be to use TMDB "Account" API calls (Favourites, Watchlist)
But I´m not the one to decide on that, if you feel it helps then go for it. This shouldnt sound negative because this add-on in general is really neat.

[removed151214]

Negativity isn't a problem - not that you're coming across negatively - it's often the negative feedback that really helps shape a finished product.

I do genuinely think the TMDb Keywords help - they're weighted lowly at the moment because there are some valueless keywords that come up consistently (there's one, for example, for all movies which have a little stinger after the credits) which I need to look into filtering out, and because I haven't adjusted their default weighting after making changes so that groups (like Keywords, Genres) etc don't give higher weightings just because there's more of them.

The related, on the other hand, will remain weighted lowly by default. The problem with the Favourites/Watchlist is the same as for a list of related movies - they're very hard to weigh in a useful manner, when all other weighing is done against metadata.

[phil65]

you could fetch metadata exclusively for favourites, watchlist items etc. I´m just sayin because I already saw several logs from people with >10.000 movies (which means >10.000 API calls.)
Anyway, I will leave it up to you, perhaps I´m too careful when it comes to dealing with APIs^^

[removed151214]

Interesting. My instinct is to say that doing that wouldn't affect/would have very little effect on the results that the widgets returned - unless the items in the favourites/watchlist share same keywords, they wouldn't match and so wouldn't gain any extra weight - there's also a good chance that such movies will be marked watched in Kodi's library, in which case they're ignored by the script anyway (and no extra data is grabbed from TMDb as is). But I'll have a think through the behaviours in this scenario. It may be I need to move to more of a hybrid (content + collaborative) recommendation system anyway, but not until I've at least released this one

And perhaps I'm too blasé when it comes to APIs

[mgonzales]

FYI - I saw a new commit to the GitHub so I pull down the latest master and now it's working for me. Do you have a repo that I can add for updates directly from your GitHub? Also I noticed the version number in ad addon.Xml hasn't changed but the code had.

In any event thanks for posting the update I'm glad it works for me now! ;-)

[RockerC]

Any chance of getting even smarter with discover recommendations of "similar" Movies and TV Shows scraping that recommendation data from IMDb?

IMDb.com have this as a "People who liked this also liked..." section which simply recommend similar movies, for direct link add "/recommendations":

http://www.imdb.com/title/tt0110912/recommendations

http://www.imdb.com/help/show_leaf?perso...=tt_rec_lm


TheMovieDB.org also have a section for "Similar TV Shows" (and "Similar Movies" for movies too but it is not as good) though not sure if it is available via its API

https://www.themoviedb.org/tv/1972-battlestar-galactica


There are also other services online that can give you recommendations for Movies, TV Shows, and Music Artists or Songs based on similar taste

http://www.similarkind.com
http://www.tastekid.com
http://www.jinni.com


This could help users discover similar media that are also in your databases in Kodi based on mood too, "If you enjoyed this title, maybe you will also like..."


I guess best thing would be if Kodi would actually scrape that info into its database and the widget would cross-check it with other media in your collection.


MoreTacksLikeThis.com also get music recommendations from Spotify that is lind of like the old iTunes "Genius Playlist" feature
http://moretrackslikethis.com

From IMDb - No. The api simply isn't good enough (and essentially non-existent for tv series.)

From TMDb - some can already be enabled from the scripts settings ('Similar' and 'Keywords') - oh, and TMDb's Similar is only good for those to series where it has this information, which is primarily only US to shows

From others - I'm already looking into Trakt, and will look into others BUT this is a content-based recommendation system, not a collaborative one. That means it bases its results on the media metadata (genre, actors, et al), not similar movies that others have enjoyed. Whilst such information can be used, it's hard to weight names of similar shows, against multiple genres, actors, etc, so any additional info will only be added if it genuinely improves the results.

Would it not be better to make the scraping part be done by Kodi's existing scrapers, so that they scrape just the IMDb and TheMovieDB tags for "recommendations" and "similar" movies into Kodi's standard video database, meaning at the same time when they do their normal scraping for all the movie metadata via the available APIs?

That way not only this widget could use the information but other python apps as well or skinners could even use that information directly from their skins without having to use widgets, or?

That is, would it not be simpler if the data for "similar" movies was always available in Kodi's video database?

So for example for the entry of Pulp Fiction, which have IMDb tag tt0110912 it would have a new additional table that also stores the ID tags of "similar" and "recommended" movies from IMDb, again see:

http://www.imdb.com/title/tt0110912/recommendations

So for the Pulp Fiction entry in the database that entry would have a child table that stored the IMDb id tags of those recommendations from imdb.com

tt0910936
tt0317248
tt0407887
tt0985694
tt0105236
tt0105236
tt0172156
tt0086148
tt0375912
tt0375912
tt0120784

That way you can with just that IMDb ID tag then do a SQL lookup in the Kodi video database to see if it have any of those movies in your database, and if do have them all it would recommend these as similar movies:

Pineapple Express (2008)
City of God (2002)
The Departed (2006)
Machete (2010)
Reservoir Dogs (1992)
Bad Boys II (2003)
The Professional: Golgo 13 (1983)
Layer Cake (2004)
Scarface (1983)
Payback (1999/I)


Of course since you might not have all those movies in your database it would be nice it the scrapers also collected this information from other sources too, like TheMovieDB.org and the others I mentioned.

Again http://www.similarkind.com and http://www.tastekid.com are possible two more good sources for other types of media such as TV Shows and music artists.

[removed151214]

FYI - I saw a new commit to the GitHub so I pull down the latest master and now it's working for me. Do you have a repo that I can add for updates directly from your GitHub? Also I noticed the version number in ad addon.Xml hasn't changed but the code had.

In any event thanks for posting the update I'm glad it works for me now! ;-)

I did post that I found the issue, but it's not surprising it got buried in the discussion over the api's. Still, glad its working for you. I'll look into a development repo, especially as I'm hopeful that most of the larger development issues will be sorted sooner than later, but won't start bumping the version on each update until such time as that happens.

Would it not be better to make the scraping part be done by Kodi's existing scrapers, so that they scrape just the IMDb and TheMovieDB tags for "recommendations" and "similar" movies into Kodi's standard video database, meaning at the same time when they do their normal scraping for all the movie metadata via the available APIs?

<snip>

Whilst it may be better for such features to be part of the scrapers this thread isn't the place to discuss that, not least because you're unlikely to attract the attention of the scraper writers here (or the core developers who would have to extend the database).

[mgonzales]

FYI - I saw a new commit to the GitHub so I pull down the latest master and now it's working for me. Do you have a repo that I can add for updates directly from your GitHub? Also I noticed the version number in ad addon.Xml hasn't changed but the code had.

In any event thanks for posting the update I'm glad it works for me now! ;-)

I did post that I found the issue, but it's not surprising it got buried in the discussion over the api's. Still, glad its working for you. I'll look into a development repo, especially as I'm hopeful that most of the larger development issues will be sorted sooner than later, but won't start bumping the version on each update until such time as that happens.

Oops sorry I just saw that reply ;-)

Hehehe

Thanks for fixing it - I really like this new widget service - have only tried the movies widget so far - but will import my tv shows on my test box and check out the TV widget soon.

Thanks!

[removed151214]

Thanks to everyone who has used this add-on, but please note this add-on is no longer supported and will shortly be removed from git.

[mgonzales]

No! :-(

It's horrible that you are leaving xbmc / Kodi development because of these Kodi team members who have made you feel unwelcome or unwanted.

I saw your other post requesting your account removed and deleted and although I don't know the exact reasons for you leaving but I am grateful for all you have done and sad to see you go.

whomever these folks are - they have to be a tiny meaningless (yet obviously loud and irritating) minority compared to the majority who welcome and have enjoyed the scripts you have contributed.

It's simply awful to see bullies who don't get their way badger those who contribute to the community until they leave.

Maybe instead of complaining and attacking they should do something helpful and contribute to the community instead of erode it.

Bottom line - It is sad to see you leave and I appreciate your hard work - skin shortcuts have made a huge impact in a good way and I would hope you reconsider and keep working on it for us grateful users if even out of your own 3rd party repo.

-Mario

[Martijn]

^^ that

[Montellese]

Uhm wait... WHAT?
Although I haven't posted in this thread I've been following it with very much interest. Would be really sad to see you go.

[phil65]

Uhm wait... WHAT?
Although I haven't posted in this thread I've been following it with very much interest. Would be really sad to see you go.

Can only agree. Was great to have you around here, Unfledged. Hope to see you back at some point of time, was always great talkin to you, no matter if via the forums or directly via PMs.

[mgonzales]

His skin shortcuts have been a godsend for the skins that use it - allowing features I could now not live without.

I doubt any of these bullies that are the root cause of him leaving are going to contribute anything nearly as useful as Unfledged has.

Sigh :-(