HomeGroupsTalkZeitgeist

Automatic recommendations

From WikiThing

Jump to: navigation, search

<<back to Help and Frequently Asked Questions

Overview

LibraryThing has several ways to suggest books you might like to read. Recommendations based on single books are available on each work's recommendations section; you can view those recommendations even if you don't have a book in Your library.

If you want recommendations based on lots of books in Your library, then you'll appreciate the three types of recommendations linked from your Profile:

  • Automatic Recommendations
  • Member Recommendations
  • Your Unsuggester

Although you can access your full list of recommendations from your Profile, new additions to your Automatic Recommendations and Member Recommendations lists also show up on your Home page.

This help page will discuss the computer-generated Automatic Recommendations, the default view.

Member Recommendations gathers in one place recommendations LibraryThing members have made for any of the books in Your library (and selected other Collections). More on this feature may be found here.

Your Unsuggester looks at your library and provides a list of books that probably wouldn't strike your fancy. If you'd like to shake up your world and try something really different, Your Unsuggester is a great place to start! More on Your Unsuggester may be found here.


Concepts

Automatic Recommendations compares Your library (and selected other Collections) to the thousands of other member libraries on LibraryThing and provides a list of some books you might find interesting. Unless you use the Collections feature to organize your books, all of the books you add to LibraryThing are added to Your library and will be used to generate Automatic Recommendations. If you use other Collections, you can choose with a checkbox whether or not to include those books in Automatic Recommendations. See the Collections page for more information on how to include or exclude a Collection from Recommendations.

What are these recommendations based on?

  • books that have the same subjects as other books in Your library (including tags and Library of Congress subject headings)
  • books in the libraries of other LibraryThing members who share books with you

By the magic of LibraryThing, you can get recommendations even if you don't tag your own books. There's an important caveat to make, though. Recommendations are based on similar subjects or tags, but tags don't always give clues as to whether you will agree with a viewpoint in a book. For instance, books about a specific religion might praise it, or they might criticize it. You might even be recommended books you find offensive. What the recommendations feature can do, however, is to help you discover books about subjects you've found interesting, because you already have books about the same topic.

Recommendations can't take into account (by ratings) whether you liked a book in Your library. If you don't want all of your books to generate Recommendations, the best thing to do is to create a special Collection; see the notes section for more information. Also see the notes section for more information about how Automatic Recommendations are generated.


Features

The parts of the Automatic Recommendations page and how they work are explained below.

Recommendations List (left side of the page)

Your 1,000 recommendations take the format:

Rank(#). Title by Author. Number of copies. Number of reviews. Average rating. No thanks! Why?

Title

Click on the title to go to the respective work page. You can see how other people have rated and reviewed the book there.

Number of copies

How many copies LibraryThing members own

Number of reviews

Number of reviews written by LibraryThing members (doesn't include links to published reviews)

Average rating

The average rating (5 is the highest) given to the book. Not all members rate their books, so if you want to know more about how many people rated the book and the spread of ratings, click on the title to go to the work page.

Author

Click on the author's name to find out more about an author and what else he/she wrote.

No thanks!

If you simply can't stand to see a specific recommendation any more, click on No thanks! and it will be whisked away. Clicking on No thanks! will NOT change any future recommendations.

Why?

Click on "Why?" to better understand why the computer thinks you'll find a book interesting. You can see what books you have prompted the recommendation. You can click any of these titles to link to its work page. Sometimes it may not be obvious why the books are linked, but members who read similar books are often similar in subtle ways.

Recent recommendations (box at top right of page)

New books, LibraryThing members, and tags are being added every day. These can generate new recommendations. You can use the links in the Recent recommendations box to "Show new recommendations" or to view your list of 1,000 "Recommendations by date."

Top 1,000 recommendations (box at right of page)

How would you like to get a set of recommendations that only covers certain topics in Your library? "Filter by other member's tags" is a great way to get recommendations for a subset of your books. Click on "Show tags" to display some of the most common tags used to describe books in Your library. If you click on one of the listed tags, the display on the left will filter to show only those recommendations related to the tag you clicked on. Click another tag to change subjects. You can click on "Hide tags" to do just that, but if you want to see the full list of recommendations on all topics, just click "Show top 1,000 recommendations."

Other recommendation lists

Another way to get customized, topic-based recommendations is to use the "Books with your tags" option. Select the "show your tags" link to display the tags you've used. You can click on a tag, and LibraryThing will use the set of books you tagged that way to filter the recommendation list. Note that the list is based on how you've used the tag for books in your library, not how anyone else may have used the tag. This feature is a great way to generate a specialized reading list based on a narrow subset of your books. One limitation, however: you must have used a tag at least 5 times to use this feature.


Notes and Frequently-Asked Questions

How does the computer calculate Automatic Recommendations?

While the specific method is a closely-guarded secret, the computer algorithm generates recommendations based on patterns of book co-occurrence in libraries, subject headings, and tags. There are rumours that the Early Reviewer algorithm accepts brie, but the Automatic Recommendations algorithm cannot be so easily swayed!

Warning: statistics are involved! For example, let's say there are 1000 LT members. 500 of them have book X, and 100 of them have book Y. If these two books were unrelated and entirely randomly distributed, you would expect 50 members to have both. That is, if everything is random, 1/2 of any subset of LT members should have book X, so if the subset is "members with book Y", then we would expect the overlap to be 1/2 * 100 = 50. If significantly more than 50 people own both X and Y, that indicates that ownership is not random, and that people interested in one book would be interested in another.

This algorithm also takes into account similar subject headings and tags for books in Your library. It uses subject headings assigned by libraries and the most common tags applied to your books by all LibraryThing members, so even if you don't tag, you can still generate recommendations. The computer adjusts recommendations (weights) based on tag/subject obscurity - so the fact that two books share the widely-used tag "fiction" counts less towards making them similar than if they share the tag "books about books" or "dystopian steampunk". If you share a lot of unusual books with another LT member, you may find that the algorithm suggests other items from these similar libraries.

When the computer generates your recommendations, it has to make these calculations for all of the books in your library compared to all other books - which is why it can take a while to load!

Why don't recommendations take rating into effect?

Doing book-to-book, library-to-library comparisons to get recommendations is already computationally expensive. Adding ratings to the mix would exponentially increase the number of calculations, and hence the time, that the recommendations engine has to do its job. Additionally, only about 10% of books on LibraryThing have ratings - either people haven't read them yet, don't remember them well enough to rate them, or just don't bother rating any books in their catalogs. The idea of generating recommendations based on ratings is out there, but is unlikely to be implemented any time soon due to the computer processing power it requires.

How can I help the computer generate better recommendations?

Recommendations are based on what books you own or have read and what subjects you find interesting. This suggests two ways to improve recommendations: include more books and more subjects. How do you do that? It may sound obvious, but recommendations will be much better once you've entered all of your library, so finish entering your books! If there are books that are your favorites, but you don't consider them part of Your library, you can add them to your Read but Not Owned Collection; some people do this for books they check out from the public library.

Creative use of tags is limited only by your imagination, and unique tags can help those with special interests find books. This can expand beyond just the subjects in a book. For instance, J. R. R. Tolkien and C. S. Lewis were in a social/literary club called the "Inklings," and some people use this tag. Maybe you have a passion for "African-American science fiction." With as many LibraryThing members as there are, unusual tags can spread and generate interesting results.

How can I stop getting all these Star Wars/Comic/Children's/Unwanted Topic recommendations?!

To keep certain types of books from overwhelming your recommendations list, create a new Collection--you can call it recommendations or some other name--and add only the desirable books to that collection. Check the "include in recommendations" box for that new Collection. Then change the Your library collection settings to remove the "use for recommendations" checkmark. See Collections for more information.

Why do I get recommendations for books I already have?

There are two main reasons you might get recommendations for books you have listed in LibraryThing.

  1. The first possible reason is what's known as the Part/Whole problem. The system can't recognize that a box set or an omnibus edition includes all of the parts within it. This work-within-a-work problem will be fixed in a later improvement to LibraryThing.
  2. The second possible reason is that the record for your book hasn't properly combined with other copies of the same work. If you don't feel comfortable tackling a work combination, the Combiners! group can help you sort this out.
Help/FAQs | About | Privacy/Terms | Blog | Contact | APIs | WikiThing | Common Knowledge | Legacy Libraries | Early Reviewers | 89,467,380 books! | Top bar: Always visible