Comparing recommendations

My husband has been playing with some information from WFG (and thinking about how to make participation more useful), and has put up a tool for comparing users' recommendations. Go to and enter your user-id (the last part of the URL for your shelves).

The results show the amount of overlap in recommended listings you have with other users. Following the "details" links shows which listings you agree and disagree on. The listings that others have recommended that you haven't might show something you'd like.

Trying this for me (s-d-youngren), shows that I have one recommendation in common with each of five other users: shutsumon, meilin-miranda, m-e-traylor, ejames, and dash. Checking the details shows that it's the same recommendation for all of them (The Aphorisms of Kherishdar), and I notice that two listings come up in their recommendations more than once: The Astonishing Adventures of Lord Likely and Ember. So I might find those especially worth checking.

Mark is still looking at what might be useful, so if you try this out, we'd be very curious to know what you think about it.

One important note: The data are not real time. If you make any new recommendations, they won't show up unless Mark downloads and processes the listings again.

Have fun.


OMG this is so fun! Miladysa, you are my closest recommendations twin. We have 5 recommendations in common!

I realize this is an old post, but I have to say, I think that kind of thing is ultimately going to be a more useful tool than averaged numerical grade reviews. Successful weblit is almost always going to be niche. The winners aren't people who try to compete for the shiny brass ring; they're people who filter gold from a muddy creekbed. This is not to say reviews are useless; the written reviews are very nice in giving someone an impression of what a story contains that may bother them or may appeal to them, but whether a given story is given a 5 star review or a 1 star review by -most people- doesn't begin to enter into the question of whether or not it can be a self-sustaining successful venture.

When you're an independent author, no amount of people hating a story will drive "you out of business", and it only takes a relatively (to traditional standards) small amount of people loving your work to keep you in business.

As this site grows in popularity, the star reviews are going to become more meaningless and less useful. You write a furry story? Too bad for you that the internet at large doesn't like furries. The furry community, I'm sure, LOVES a well-done furry story. If you write a good furry story and you got a million random people to try your site (we're assuming you have unlimited bandwidth and processing capacity) and review it, your reviews would be awful... but you would be set for life, because you'd have gained thousands of true and loyal and supportive fans.

So, here are some thoughts: if there are people in the community who are good at developing these sorts of things, can we possibly get user profiles here tagged with interests? And then when we like a story, can we have "Readers who like this story are interested in ________" and "Readers who are interested in ________ like the following the stories"? Because that would be useful. And it wouldn't have to be limited to things like topics and types of mature content and genres. It could be things like "vivid description" and "introspective narrative".

Somebody who likes vivid visuals would not like my writing. Somebody who likes introspective narrative might.

I would rather people have tools like that than know that Tales of MU gets X.Y stars.

Success in weblit is not about writing a story that everybody likes. It's about writing a story and getting the attention of people who will like it.

Hi Alexandra,

At the moment, we do show recommendations in common on each listings (ie. "people who liked this also liked") -- which are counted on recommendations in common, so it should provide reasonably relevant answers. But, yes, it could be better. I'm currently brainstorming ideas on the next version of the WFG software, and I think I'm going to vastly "decentralize" the design -- to try to de-emphasize ratings in favour of currency and connections. At the moment, the design relies on a very active editorial presence, which has proven largely unsustainable in the long term. I'm hoping to find something that relies less on that, while still keeping things from becoming purely a popularity contest.

Please do continue to post if you have any ideas. And that goes for everyone.


Oh, yes, the recommendations in common is one of my favorite things the site has going on right now, and I like the detailed reviews. Anything that avoids a popularity contest is good in my mind, because not only does a popular ranking system encourage a "rich-get-richer" system at the expense of newbies, but it also erases the central strength of weblit, which is that there's something for everybody. 1 star story might be the infamous "crap" of Sturgeon's Law but it might just not be a popular subject or format.


Oh. Oh.

I just figured out how to do what I was talking about, but simpler. I said having readers tag themselves and then try to figure out what kind of person likes what kind of story. That's really complicated, though.

But what if readers/users/reviewers can tag a story? Like how you can tag products on Amazon. If I like a story or don't like a story but the reason is because it's heavy on narration, the fact that it's heavy on narration is useful information to everybody else who either likes that sort of thing or doesn't, right? So if I could put a stamp on the story and then people click on "narration-heavy" and they see a list: 17 people tagged Webbernet Tales narration-heavy, 11 people tagged Adventures on the Internet narration-heavy, and so on.

There should probably be a pool of tags so that people aren't going around tagging stories with descriptors like "shit" and "waste of time" (because these are subjective judgments); an editor would be needed to approve add new tags by user request. It could cover moods (light, dark, serious, funny, introspective, angry), subject matter and genre, style, and things like frequency of update and consistency of updates, so that people can find what they're looking for and know what they're getting into.

Are these 'tags, recommendations and reviews' similar to those Chris has begun putting in place at Muse's web fiction wiki?

Interesting ideas. I wonder if readers are that self-aware or consistent about what they like in a story? I'm not sure I am.

Fiona, it seems to me that if it was a tagging system, I wouldn't need to be consistent as a reader. I could tag my own profile with 'introspective narrative,' 'tons of explosions,' and 'travel diary,' all of which I happen to love but which tend to be mutually exclusive. Or search by any one of those.

Letitia, that's a great place to start.

I suppose I'm thinking of like a combination of Amazon's user-driven tagging and Pandora's "music genome", especially when I mention tagging things for being narration heavy or introspective or whatever. I don't think there are a lot of story sites that categorize things on those bases, but I feel like it would generate some interesting lists and comparisons.

Muse's Success, another listing service, has a very extensive tagging system, just as a relevant example:

I think I need a little more information on what you guys are asking for, as we already have very extensive tags here on WFG: -- the most common tags are listed across the top, and the "more" link shows the full list. Every listing also shows the assigned tags, and clicking through on them takes you to the list of all listings with that tag. You can combine tags, and the recommendations shown will be constrained to the selected tags. Is that functionality just not clear, or are you asking for something different?


There are two things I'm looking for, one of which might be there but isn't obvious and the other one is more a matter of culture/practices than coding.

The first one is reader assigned tagging. Is that there? How do we do it? When I look at an entry I don't see a "tag this entry" link. Ideally you wouldn't have to review a story to tag it, as that would restrict tagging to the smaller number of people who are willing to sit down and write a review, or result in more people writing shorter/shallower reviews just to tag.

And the reason I want reader assigned tagging is because of the second thing I'm looking for. See, the extensive" tagging system in place now is much like the D&D two axis alignment system: an interesting intellectual exercise in categorization that doesn't tell us as much as we think it does.

People say "I like science fiction" or "I like horror", and they mean it. And having tags for those genres helps narrow things down to the point where they can start looking for things that they'll enjoy. But within the category of science fiction, there will be certain things -about- a given story that are the real things someone likes about them.

To give a specific example: imagine there were tags for "fantasy" and "school of magic". "School of magic" is already more specific than most of the tags I see on the site. But if we're relying on those tags, then Harry Potter is the same kind of story as Tales of MU, and one of the banes of my existence is people coming in and saying, "Wow I thought I'd love this because I love Harry Potter, but this is nothing like Harry Potter." or "You should probably be more like Harry Potter. I'd like this story more."

I'm looking for a more extensive set of user-applied tags that deal with matters of style, more specific sub-genres, settings, tropes, content (as in "mature", "explicit", "fetish"), etc., so we can start getting real data on what people actually like/dislike about stories, and ways to find stories (even in genres we don't normally read) that might appeal to us.

One of the advantages of this is it would cut right through the weakness of the "People who recommended this also recommend..." system that will become more apparent as the site grows, and that is, that only works for connecting stories where readers of one are already reading the other. This would let give so many different ways to find stories: this story is "most like" (shares the most tags) that story, these stories all have this tag, etc.

Yeah, letting readers put them in (instead of authors requesting them from Chris) might be a good way to get them more precise, and give people handy references for what they might enjoy. But will a lot of readers go that far in their descriptions, given most people write very short reviews to begin with?

Some examples, for discussion:

With the first two, there is less than one page of listings. For the third, there are 5 pages of listings, but the tag toolbar shows the most common other tags with urban fantasy. Admittedly, the first two are pretty useless (fantasy and online novels), but some of the other top tags include "magic", "action", and "dark fantasy". If you open the "more" tag, you can find a good number of other tags in common, too.

I can certainly see adding mood ("dark", "funny", etc.) as tags -- which is largely about encouraging authors to choose them when listing. And I don't have any objection to allowing users to tag their own collections. But, as most users don't bother rating more than one story in the entire collection, I think Gavin's got a real point: giving people more opportunities to be heard isn't necessarily going to improve things, as they aren't taking the opportunities they already have.

One of the largest problems I've encountered, recently, on returning to more active administration of the site, is just how many listings have completed, stopped being free, or disappeared entirely. Within the first three pages of the uncategorized listings, for instance, I had to change the status on more the half the listings. That's something I really want to address in the software -- possibly by considering recent RSS activity, recent ratings, and TWF votes count for marking things as current.

I also want to build in more incentive for people to share their readership. One of the reasons TWF is successful as a traffic source is that, when people vote, they are generally already finished reading the thing they love, and so might be willing to try out other things. It gives popular authors incentive to share their readers with less popular authors. That's something I think I can address in the redesign, something that people have clearly been unwilling or unable to do with the current structure. As I mentioned in another thread, there really are good reasons for people to contribute back to the commons. Encouraging them to do so seems like a high priority, to me.

Don't get me wrong -- I'm definitely open to ideas on how to improve things. But I'm wary of any ideas that rely on more reader involvement with a substantially similar design -- because I just haven't seen any evidence that readers will rise to the occasion. :-(


Just an aside to main topic here, but Chris, why are you counting "completed" as a problem? Just means the readers have a complete story to read instead of waiting for updates, right? Some authors prefer to post complete novels in the first place.

I would say the three problematic issues are "stopped being free", "disappeared entirely", and "abandoned mid-story". Unfortunately the latter is hard to distinguish from a very extended hiatus.

Getting back to main topic, I wonder if maybe the existing tagging system just needs to be more obvious. I wonder how many readers notice it and figure out that they can search on multiple tags. Maybe something that looks more like a conventional Search function?

The being-heard-is-not-enough thing is partly where we came in. A catch phrase for our thinking might be: "My contributions help me too." If I rate a review or recommend a story, this should give me more information on whom to trust and what to read.

Chris, aside from the tags, Mark has a wish-list item: A way for users to "mark as read" listings without having to rate, review or recommend them. This would be useful for filtering, but he's really after an unambiguous way to detect "chose not to recommend". Right now we can tell if users agree on liking a story, but not if they agree about disliking it. Having that would allow the method we described in the OP to do more detailed matching of people. (As with YouTube's "thumbs up / thumbs down" compared to Facebook's "like"; more complete information.)

Oh, and I agree with Fiona about not downplaying completed stories.


I'm not convinced that "chose not to comment" unambiguously equals "chose not to recommend". It seems common over all the web writing communities that relying on reader participation is soul-destroying. The general lack of input is as difficult to interpret as a dozen short, single comment 5 star reviews in the current system.

"Don't get me wrong -- I'm definitely open to ideas on how to improve things. But I'm wary of any ideas that rely on more reader involvement with a substantially similar design -- because I just haven't seen any evidence that readers will rise to the occasion. :-("

That is a common experience. Habits change slowly and for centuries the bookbuying public have taken their book home and said no more about it to anyone. A dozen international reviewers are respected, a few locals too, but the majority of readers simply do not feel inclined to comment. Especially after they have completed their read and moved on to the next.

Hi guys,

First up, on marking stories complete. Imagine what the NYT Bestsellers list would look like if it was cumulative, instead of current. Harry Potter book 1 would probably sit at the top forever, followed by a number of other famous books. Stuff that's selling well today would never even make the bottom of the list. The listings take into account currency because otherwise the top of the list would fill up with stuff that's long finished. So, once marked complete, the system starts ageing them down the list.

On that front, what I would like to change in the next version is to take into account reader activity, rather than writer activity. Just because something is complete or abandoned doesn't make it irrelevant to readers -- I've just been using the information I have to approximate relevance. I think I can do better in the next version.

Shelley: I'll add the "have read" to the list of options already there. That, or I'll allow readers to mark a listing "not recommended", which might be easier to present. I keep being tempted to get rid of ratings altogether, in favour of a thumbs up/thumbs down system, but I can't help worry about the loss of detail that would entail.


Oh, and searching for a tag does work.