Where's the Pandora for e-book recommendations?

[rahlquist]

Originally Posted by Slite

Which IMHO is a very very bad idea. Whats going to stop the makers of that reader to "abuse" that kind of data by planting ratings/recomendations? Or if not that, using it as a guideline for selecting what kind of author/genre they should supply?

Which is why as part of the recommendation engine you never expose the underlying internals. Pandora works much better the more data you feed it (least it does for me, rarely do I get a bad song from it), it seems almost Bayesian in its learning. Some yutz giving a bunch of phony 5 star ratings for one entity isnt going to completely fool a smart recommendation engine. However thinks like Amazons ratings system being used as a recommendation are a joke. How many times in the past 4 years have people been exposed for paying people for Amazon reviews? Whereas adding the social point to a hardware ebook device allows the ultimate in control for someone building the backed. You only get one vote.

Just some thoughts...

[BearMountainBooks]

And not just Amazon--Kirkus does (or did) accepted books that they agreed to review--if they were paid something like 400 dollars (using their reputation--just getting a Kirkus review was something some people thought meant quality.)

But that said, I do find the Amazon reviews useful if they say what they liked or didn't like. Especially if I'm not sure what the book is about. For instance if 4 reviewers mention the "romance" for a book I thought was a mystery that tells me a lot more than the number of "stars." So they are useful. At least to me.

And as a writer, I'm *still* waiting for my review from Harriet Klausner. I think Amazon should give her a Kindle so that she can get going on Kindle books!!!!

Maria

[Elfwreck]

whatshouldireadnext.com keeps giving me "Error 500: Internal Server Error" results, which I suspect means "that book is not in our database."

Also, entering book 1 of a trilogy doesn't get a suggested read of book 2, which strikes me as annoying.

I suppose it'll work better when it gets more titles.

[Robotech_Master]

I'm surprised the thread has gone this far and nobody's mentioned Alexandria Digital Literature (Alexlit) yet. Alexlit was one of the first e-book sites (though it concentrated on selling e-stories rather than full-length e-books, since back then nobody was sure whether people would actually pay to read full-length e-books).

Its big gimmick was that it used collaborative filtering to let readers rate the books they liked, then it would generate a list of books readers hadn't rated that they were likely to like based on what other people with similar tastes enjoyed. In other words, exactly the kind of thing you're looking for.

When it worked, it was absolutely brilliant, and it introduced me to many of my now-favorite writers who I probably never would have discovered on my own. Sharon Lee & Steve Miller, P.C. Hodgell, Vernor Vinge, Janet Kagan…

Sadly, it went down some years back and has never come fully back. They were planning a return in 2008, but then the stock market imploded and capital dried up.

Hopefully it'll still happen someday.

I interviewed Dave Howell, AlexLit's creator, a couple years back in my Biblio File podcast. Interesting stuff.

[Phogg]

Originally Posted by AMacD

Some books are so poorly written that it takes less than a paragraph to reject that one. Authors that don't know the difference amongst "there, they're and their" for example. To is not equal to two or too. Or foul language just for shock effect that doesn't lend to the tone of the story. It should not take a quarter of a book to establish a love of the characters. I finish most books that I have started and managed to get through the first quarter without losing interest or getting so confused that I am totally lost. On the other hand I am not going to waste my time on a hopeless book when there are a thousand more in the queue waiting to be enjoyed.

Good point.

There are some exceptions though.

Walter Gibson (Under the pen name Maxwell Grant) intentionally spelled the word "clue" as "clew" throughout the famous Shadow pulp novels.
But those are a good read.

[Robotech_Master]

Originally Posted by Bob Russell

My thought for today is that something similar should work great for books as well. Study what makes books worth reading and fun or valuable. Ask people what books they like and don't like, and why. Provide a catalog of book qualities and use it to provide a personalized stream of book recommendations that can be evolved with your additional input about books you've read. In other words, do something a little deeper than Netflix or the standard social book recommendation sites. It's all the more important for books than songs - you can waste a lot more time and effort on a bad book than a bad song.

Actually, you don't need to get this complicated.

Alexlit, which I mentioned a couple posts up from here, didn't have any of that.

All that "Hypatia," the recommendation librarian software, had was a simple five-level scale, where readers could rate how well they liked or how much they hated a book. Alexlit was completely genre-agnostic. All it did was churn the database to build profiles for each person, figure out who the other people that had the closest tastes in terms of likes and dislikes to that person were, and find what books that other person loved that the one person hadn't read yet. Once it built up a big enough sample database of users' likes and dislikes, it got to be pretty darned accurate.

You don't need a catalog of book "qualities". The "qualities" of books people like is encapsulated in their rankings. They'll rank books that have the qualities they like higher, and rank the ones without them lower. So will other people with similar tastes—so Hypatia doesn't need to know whether a book is science-fiction or hard-boiled detective noir or historical romance, just that your closest neighbors all loved it so you probably will, too.

Even if you read nothing but science-fiction and won't touch historical romance with a ten-foot pole, Hypatia may well dig up a historical romance you would like despite yourself—which you would never have discovered otherwise since you don't read those books.

Countless times I've been recommended books by Hypatia that I didn't want to read because I thought they didn't look interesting. Every single time I've gone ahead and read them after all, I've discovered a new favorite book.

I miss Hypatia so much.

I've heard from Dave Howell that he's working on reconstructing Hypatia in his spare time. He doesn't know how long it will take him, but sooner or later it will be back, and you'll all see how good its recommendations are.

[13xforever]

Originally Posted by Superlucky

I went and registered so I could try it, but all I found on the site were forums. If there's a link to the actual tool I couldn't find it.

After logging in, you see 3 links on your left: BookLamp interface, Forums and Request a Book.
1) It's a matching interface where you're selecting a book and it shows you recommendations based on it.
2) Well, it's forums
3) Here you can add new books to the database.

What I don't like is that all I can do is to add a general information of the book, but can't influence a matching criteria. So, instead of looking into my book collection, finding the basis off of which you could predict what I'll like, they just making a chain of "similar books", which isn't a really good approach, IMO.
And of course, you could've noticed a shortage in the database entries.

[Superlucky]

Originally Posted by 13xforever

After logging in, you see 3 links on your left: BookLamp interface, Forums and Request a Book.
1) It's a matching interface where you're selecting a book and it shows you recommendations based on it.
2) Well, it's forums
3) Here you can add new books to the database.

Thanks! I'm really not that dense, but when I first signed up, logging in took me directly to the forums page for some reason. Even now, if I navigate back to the main page (while logged in), the "Click here to enter BookLamp.org" link goes to the forums page and I have no idea how to get back to the page you referred to without logging out and logging back in. The "Home" link even redirects to the forums. I do not like the site at all.

[Robotech_Master]

As a side note, Alexlit is coming back soon. (It "looks" back now if you go to the site, and it produced a new list of recommendations for me, but it still has lots of show-stopper bugs that will probably keep it from working very well for new subscribers.)

I'll let people know when it's more "ready for prime-time."

[AMacD]

Originally Posted by Phogg

Good point.

There are some exceptions though.

Walter Gibson (Under the pen name Maxwell Grant) intentionally spelled the word "clue" as "clew" throughout the famous Shadow pulp novels.
But those are a good read.

I noticed also that Edgar Rice Burroughs writing in the '20's used "clew" as well. I wonder if it was in wide use in those days? It is before my time. He was very consistent in its usage.

[Robotech_Master]

Originally Posted by AMacD

I noticed also that Edgar Rice Burroughs writing in the '20's used "clew" as well. I wonder if it was in wide use in those days? It is before my time. He was very consistent in its usage.

I believe so. There were a lot of words spelled "funny" in those days. For instance, "canyon" used to be spelled "cañon".

[ThomasMc]

Good luck signing up with BookLamp. I tried about 20 times, but it would never accept the stupid "CAPTCHA" phrase. (and yes I used the correct case).

[Robotech_Master]

For what it's worth, Alexlit seems to be up and running now. Still a bit flaky from time to time, and currently it's limited to no more than 5 simultaneous users at a time until Dave Howell finishes rewriting it for a new database back-end (so it might be best to wait for the early morning hours when fewer people would be using it).

But if you're looking for a "Pandora" for e-books, you won't find anything closer than Alexlit.