Looking for the least common denominator: Merged receommendations from The Storygraph

[Desertway]

Originally Posted by hildea

so the machine learning is looking at our to-read-piles instead of (or in addition to?) our reading history. At first glance, that seems reasonable, but I tend to be more ambitious when planning what future me should read than when deciding what current me is going to read So I have a lot more non-fiction books and heavy-ish books in my to-read-pile than in my reading history.

I see. Yes, that way of finding recommendations is not going to work that well for me either. I have books on my to-read list that go way, way back and don't really reflect my reading preferences now. And I'm sloppy. Some of my to-reads are on GR/SG, and some just go right onto my Amazon wishlist.

[hildea]

Originally Posted by JSWolf

Can you do this again with my books added please? It should be possible to tell the system to ignore specific recommendations so can you ignore any that are not English? Thanks.

I did all of them with you included. Or do you mean that you added more books yesterday, so that the results should be different now?

The way it works is that I get three suggestions, and I can refresh the page to get others. There's no feature for telling the algorithm to ignore specific books.

@UncleRobin: Sure, I can do that

[Uncle Robin]

Originally Posted by hildea

@UncleRobin: Sure, I can do that

It's possibly the BIGGEST difference between TSG and GR - although TSG has only TWO tech staff (out of 3 in total), one or or other of them will answer every query like this

[JSWolf]

Originally Posted by hildea

I did all of them with you included. Or do you mean that you added more books yesterday, so that the results should be different now?

The way it works is that I get three suggestions, and I can refresh the page to get others. There's no feature for telling the algorithm to ignore specific books.

@UncleRobin: Sure, I can do that

With regular recommendations, you can remove books from being shown. So I thought maybe it's the same with the Plus buddy read recommendations.

Thanks for doing the recommendations. Some odd results (IMHO). I am surprised that there are no mystery books as I do have a lot of mystery type books in my list as well as SF and fantasy.

[issybird]

Originally Posted by hildea

so the machine learning is looking at our to-read-piles instead of (or in addition to?) our reading history. At first glance, that seems reasonable, but I tend to be more ambitious when planning what future me should read than when deciding what current me is going to read So I have a lot more non-fiction books and heavy-ish books in my to-read-pile than in my reading history.

<SNIP>

I guess I should suggest that they use reading history instead for the buddy read suggestions. Unless I'm atypical in being overly ambitious on behalf of future me...

Disclosure: I’m StoryGraph curious, but every time I think of porting my data from GR and getting up to speed I feel tired.

That said, using wishlists for anything at all predictive can’t go well. One of my pet peeves at GR and it’s down to the users, not the site, are people’s overwhelming WTR lists. I have friends with WTR lists in the several thousands, even in the five figures, which is ridiculous. I know the GR mantra is that people use the site to their own ends and generally I’m on board, but what possible use is a wishlist of such gargantuan length? And annoying to me to the extent that when I try to compare my own books to someone else’s wishlist books are included. Clearly it should only be books that have actually been read and rated.

StoryGraph really needs to rethink this.

[astrangerhere]

Originally Posted by issybird

Disclosure: I’m StoryGraph curious, but every time I think of porting my data from GR and getting up to speed I feel tired.

/SNIP/

StoryGraph really needs to rethink this.

Issy - it is very little to get up to speed with, honestly. Just come in and wander around some of our pages and see if it strikes you.

I am only getting used to it in terms of remember where to go to see a view that I was used to using on GR and small stuff like that. I never used the GR social aspects and I am more likely to talk to a librarian or indie bookseller about book recs than I am folks on a website. But I love the clean, ad-free interface and the focus just on books.

I had a few false starts before I went all-in, but I am quite pleased now.

I agree, however, with your point about where the "recommended reads" comes from. I would definitely prefer it review the categories in my read, which is in the thousands, as opposed to the 10-15 on my TBR list.

[Desertway]

Do we know if SG considers genre tags when making recommendations?

[Uncle Robin]

Originally Posted by Desertway

Do we know if SG considers genre tags when making recommendations?

I don't know if there is a difference between reading recommendations for buddy reads versus individual recommendations.

In the case of individual recs, the answer is definitely "yes" - the reading preferences survey asks for genres you like and dislike, along with many other perferences and dislikes. The individual recs are generated from the responses to that survey, not from the TBR - as can be seen in that it will generate recs with a zero-length TBR and the recs will change if the survey responses are changed, with no change to the TBR.

The caveat is that the above applies to one's own recommendations, I don't know how the buddy read recommendation algorithm works

[Desertway]

Originally Posted by Uncle Robin

I don't know if there is a difference between reading recommendations for buddy reads versus individual recommendations.

In the case of individual recs, the answer is definitely "yes" - the reading preferences survey asks for genres you like and dislike, along with many other perferences and dislikes. The individual recs are generated from the responses to that survey, not from the TBR - as can be seen in that it will generate recs with a zero-length TBR and the recs will change if the survey responses are changed, with no change to the TBR.

The caveat is that the above applies to one's one recommendations, I don't know how the buddy read recommendation algorithm works

Thanks very much for that explanation. I do find the individual recommendations are pretty spot on. As for the buddy read recommendations, I was thinking about the genre tags that TSG applies, because they aren't always the tags that I would apply. For instance, many books in my TSG lists are tagged as "fantasy", though that isn't how I would use that tag. This might result in group recommendations that don't really work. Not that I object to the way the tag is being applied, there are no hard and fast rules here.

[hildea]

Originally Posted by Desertway

I do find the individual recommendations are pretty spot on.

I agree, and since the individual recommendations are so good, it surprised me that the buddy read recommendations seem so unimaginative. I also find that if I refresh the buddy read recommendations a few times, I get a lot of repeating suggestions.

[hildea]

Some info from StoryGraph:

​Yes for Buddy Read suggestions, we look at the books on the group's to-read piles and suggest the books we think they will like the best. We thought there'd be a higher chance that people would be excited to pick things up they've already earmarked to read - but thank you for the feedback, we will certainly keep this in mind!

Individual recommendations are driven by a range of things, but the primary force is the reading preferences survey. :-)

[JSWolf]

Originally Posted by hildea

Some info from StoryGraph:

That makes sense why the recommendations don't make sense. I would rather they used the book in my list of books I've already read and gave a 3+ star rating to.

[JSWolf]

I have my own TBR list, but I don't want such on SG or GR.

So that means that for me, a buddy read recommendation on SG is out. We'd have to figure another way to do it.

[astrangerhere]

Originally Posted by JSWolf

I have my own TBR list, but I don't want such on SG or GR.

So that means that for me, a buddy read recommendation on SG is out. We'd have to figure another way to do it.

The most effective buddy reads are literally just inviting users with similar books to you. I don't view buddy reads as a book club. Its more a "I am going to read this book, would anyone like to read it with me?"

[hildea]

Originally Posted by astrangerhere

The most effective buddy reads are literally just inviting users with similar books to you. I don't view buddy reads as a book club. Its more a "I am going to read this book, would anyone like to read it with me?"

Yes, that seems much more likely to be successful.

Originally Posted by JSWolf

So that means that for me, a buddy read recommendation on SG is out. We'd have to figure another way to do it.

Heh, I guess when people have very different tastes, the old fashioned way of starting a thread and trying to find something in common works better. Or seeking out a book club with people with similar tastes.