Automatic Recommendations

What are Recommendations?

The site tries to automatically determine which kind of stories or series you like, based on:

  • On which stories you’ve clicked “Favorite”
  • Which stories you’ve rated favorably
  • Which stories you’ve read

Where "clicking Favorite” is a much stronger indicator than just reading a story, of course.

The algorithm tries to find stories

  • Using the same category as your most liked stories
  • Are written by the same Author as your most liked stories
  • Use the same tags as your most liked stories.

Using all this, the site can get an idea of what you like and what you don’t.

Of course, this is not rocket science; it’s just statistics, and of course, the algorithm can get it all wrong.

Important:
The more feedback you give us, i.e. the more your read and, most importantly, rate and use the “Favorite” button, the better the recommendations will become!

Where are the recommendations shown?

  1. On the main page, there is a tab “For You”:
    image

    In that tab, you’ll always find a random selection of stories/series we hope might be interesting for you. You can click the “Update” button to get a new random selection.

  2. In the menu under “User→Recommendations For You”
    There you’ll get a list of 100 recommended series, starting with the series the system believes fits your taste the most.

  3. If you’ve read a story, at the bottom you’ll get to see a list of 3 series which are similar to the one you’ve just read and which should fit your taste.
    BUT
    To unlock that list, you have to rate the story you’ve just read! That’s our incentive for you to rate the stories you’ve just read!
    (Pssst, don’t tell anyone, but commenting on the story or clicking “Favorite” will also unlock that list of those three recommendations!)

It’s important to note that any series you’ve already read completely will not be suggested. If you’ve read some chapters of a series, but not all, it will remains a candidate for recommendation (actually based on the tags of the chapters you haven’t read).

I really don’t like that series the system keeps recommending!

Nothing is perfect, and the algorithm can definitely get it wrong, as it’s just based on tags. No one ever claimed that every story tagged with your favorite tags is really to your taste.

But the system will keep suggesting it to you, which can be annoying, and it also takes up space for another story that should be recommended instead.

That’s why there is an option to tell the system that you don’t want a given story or series ever to be recommended to you again.

Every recommended series has this indicator in its title:

image

If you don’t want to see that series recommended, just click that golden “sparkle” indicator. It will become red, which means that this series is now blocked for you.

I hope you like this! The algorithm can and will be improved, so please let us know if you have questions or suggestions about it!

11 Likes

What a great addition! Not only we get a new way to find new and amazing stories based in our taste but readers will also get new incentives to rate, comment, or like stories.

Congratulations!

And here comes a question: Does the feature have into account the tags and categories we have marked as unwanted? Like, if I unselected Transformation stories in the main page, will they also be off in the recommendation system?

Thanks!

2 Likes

Thanks!

Blocked tags are honored everywhere, including the recommendations. I hope. I have to be honest, the variations are so mind-blowingly big that I couldn’t test every scenario. So if you see anything that doesn’t seem to make sense, just let me know and I’ll look into it!

2 Likes

So far so good! But I’ll take into account leaving a message here if I see something odd.

2 Likes

Hi! Me again!

Everything is going good, so nothing to report :+1:

It’s just that this new feature inspired a suggestion. Would it be possible to add a tab to the bookmarked stories? Somewhere with the “New Stories”, “For You”, “Top Rated” etc. I check and I’m pretty sure it fits well in both the computer and phone screens.

I believe It’d make it easier to access the user’s bookmarked stories, and it could potentially make the feature more visible to those who do not know about it :slight_smile:

1 Like

The tabs are getting a bit out of hand, tbh, I’ve already worried that this “For You” tab might be too much.

I’ll discuss and consider it. But thanks for the suggestion!

1 Like

Thanks to you for keeping it in mind and considering it :blush:

I imagine it will become more accurate as it learns but is it possible to turn off the recommendation list? I like the randomness of exploration.

I absolutely love the idea of this, but my personal results so far… lacking. I feel like I’ve done quite a thorough job of favoriting and rating over the years. And I’ve gone through the first 20 or so of my recommendations and skimmed through a bit of them… and literally none of them are remotely interesting to me. They seem like completely miscellaneous junk.

I do hope you will keep iterating on this, because there is definitely potential here. It just seems like it’s not hitting the mark yet.

2 Likes

Well, what do you mean with turning them off? If you never click that tab, you’ll never see them. It’s not like they’re forced on you.

The ones showing up after reading a story kinda do, but I think that’s not so intrsuive that it can’t be ignored, is it?

As of now ( :slight_smile: ) there is no AI involved in finding those recommended stories. It’s all based on the tags and authors, basically.

So if you’ve already read most of the stories of your favorite authors and if you’ve already read many of the stories with the tags you like, there not too much left to suggest which would be a perfect for fot you, because it will never recommend stories you’ve already read.

Note that there’s also a lot of randomness involved, so you get different suggestions each time you open the “For You” tab, which of course also make it less accurate.

If you look at the sorted list of recommended stories which is available through the menu (User → Recommendations For You), you should get a good idea what the site thinks is your taste. It’s sorted by the internal score we build based on the tags and authors you seem to like.

I will also update the user profile shortly to show the list of favorite tags and authors (currently, the profiles are not using the same algorithm as the “recommended series” approach) so you can check if this is accurate for you.

Ok, I think I’ve addressed some issues. A new version is uploaded, and I hope it’ll improve the quality of the recommendations:

  • I’ve removed several tags from being considered: the “person” (1st person, 3rd person…), the AI tag (“contains AI patterns”) and the “contains images” tag. Because they don’t actually define a user’s preference and just water down the results.
  • When looking at Authors you seem to like, it was looking at each story individually. So authors with very long series, where you liked/rated each chapter, were dramatically overhyped. I’ve scaled that so that when looking at authors, only the whole series counts, not each story. That definitely fixed my list of favorite authors, which looked very strange at first… :slight_smile:
  • Also, the check for which authors a user likes now reaches far further into the past - while the check for tags is way shorter (taste in tags might change, but your favorite Authors remain…)

You can now look into your own profile to check if the site got it right.

In your “Reading habits,” the list of “Preferred Kinks” and “Loyal fan of…” now lists the actual tags and the authors the recommendation algorithm also uses to find stories you might like.

1 Like

Cos I’ve gone under more than one author name, though they all are under the same account / email address , I’ve noticed I’ve been recommended a few of my own stories :rofl::rofl::rofl::rofl::rofl:

2 Likes

Oh. This is not intended. It uses the ACCOUNT, actually, but it doesn’t block stories you’ve authored yourself. I thought about that but decided against it, as I figured everyone has read their own stories at least once …

Maybe I was mistaken. I’ll add that!

1 Like

Aww it’s No worries at all :smiling_face: it’s a cool new feature

1 Like

Props for making this an algorithm and not resorting to ML or “AI” methods.

2 Likes

AI is very bad at a lot of things and extremely unreliable. It certainly would not be suited for this use case at all. Old fashioned algorithms win the day!

3 Likes

I’ve uploaded another update that fixes an issue with the categories not properly taken into account for recommended stories. They were also not shown correctly on the users’ profile (under “Preferred Kinks”).

Both is fixed now.

Oh no, I don’t feel forced upon. When I rate a story up pops a list of recommendations, not a one of which I would have clicked on. I’ve rated a couple of stories (Always 5’s) and more non sequitur recommendations. I just wanted to shut off the recommendations until the AI has had a chance to form proper relevant suggestions

It’s no AI… it’s just statistics based on what you’ve done on the site.

And it already should have a proper idea about what you like and what not, based on which stories you’ve read and liked in the past.

At least now, since I’ve fixed some issues which were affecting the results significantly.

1 Like