The article tries in several paragraphs to explain the experience of using it - I'd recommend instead just clicking on the Library Explorer link[1] yourself and actually experiencing it yourself.
There's something to be said for these physical-world-analogue UIs, in my opinion. Even if this is technically the same affordance as provided by old school web directory UIs, there's more of a sense of tangibility and immediacy to this kind of visual presentation, that makes it more appealing and easier to spend longer times in.
One bit of "meta" feedback from me: I love Open Library, but I get stuck every time I read one of the blog posts - because there's no link to the actual library! I always hunt for it for a while and eventually have to go up to the url and manually remove "blog." and the rest of the blog page parameters.
Maybe I'm just missing it? But the big "Open Library" icon is not clickable, and even the "about" page doesn't seem to have a link to openlibrary.org.
> Even if this is technically the same affordance as provided by old school web directory
The point is that unless they are extremely well made, the affordance and accessibility is much worse.
OTOH, I lowkey like those environments for social events. Maybe it's just that I'm somewhat introverted, but I find that having a playful/gamelike environment as opposed to a chat window makes interactions less awkward.
Do try out the "Settings" cog at the bottom of the UI -- one thing that makes this interface so powerful is you can add custom queries to transform the entire library (such as to only show kids books, text books, or biographies on any subject).
The way the openlibrary.org site is constructed seems better than the others that came before, e.g., WorldCat. Every page is not just HTML but JSON, too. Omit the slug and add .json after the ID.
As cool as this is, I think it would be worth your while to look up Clay Shirky's "Ontology is overrated" if you can find it, or watch his talk https://youtu.be/ujMgQqp8YSY?t=1118
Thank you for trying the Library Explorer :) Very interesting talk! Personally, I don't think anything there is a testament for abandoning classification trees entirely; just "common pitfalls" to avoid when building new ones. The Library Explorer is very purposefully classification-system agnostic, so if a better classification system comes along, we can import it and switch to it! And unlike a physical library, we don't have any re-shelving costs :)
One of the core design features I was aiming for with Library Explorer was that the user should never be navigating the "hierarchy;" they should be navigating the books. Forcing the user to move through categories forces them to try to understand the hierarchy, which can at times be not super user-friendly. Note that the Yahoo/Google examples in the video do just that; the user picks a class, and then sees websites. By showing them the books directly, the user "deduces" the classification tree (lots of books about physics? I must be in the physics section); the classification labels aren't really necessary for exploration.
In general, I think there are trade-offs. A classification tree is a model with flaws (just like all models). But its core benefits are (1) it's a tree; so each node has semantic siblings, parents, and children; (2) each node in the tree has a finite number of child nodes (and usually <30 child nodes; which makes it ideal for human traversal); (3) each book can be uniquely identified by the path from the root. (3) Was very important for physical librarians (classification systems were sort of like a search index for librarians), and was the cause of some of the issues described by Shirky. But online, we don't need (3), so we can kind of throw it away. The tree can be degenerate; books can (and do!) appear in multiple nodes of the classification tree, because, unlike in a physical collection, in a digital collection, books are disjointed from the classification tree itself.
Tag-clouds are _definitely_ more flexible (as Shirky said, classification trees are a _restriction_ on the tag-cloud model), but come at a cost of being harder to navigate (as flexibility usually does) and get a "big picture" idea of. I haven't seen a good example of a UI that lets you navigate tags in a way that doesn't just let you travel from one node to a neighbouring node (please post if folks have one!). Trees allow you to travel up (parent node), down (child node), and left/right (sibling nodes). Graphs you're limited to just traveling to adjacent nodes. You could try to algorithmically _deduce_ parent/child relationships though (that would be an interesting thing to try! I'm sure algorithms exist that do this). Ways of browsing graphs is definitely an interesting problem space! Open Library does also have tags (called "Subjects"), so finding a way to make them more user-browsing friendly would be great :)
> the user should never be navigating the "hierarchy;"
My first impression when I looked at the explorer for browsing was "oh, the Dewey Decimal system. I have to figure out which top-level category my interests like, then click on it, and follow the hierarchy of the system down to the books I want. That might not be the intent, but that's what the primary interface is.
I am impressed and happy to see Open Library works without Javascript. Thanks for considering the dozens of us. It's totally reasonable not to have "Explorer" work since it sounds like an application and not a document. And even in that case the links and search still show up so it falls back well.
Thanks for noticing, our primary goal is making sure a minimal experience catalog works (best we can) for folks in lower bandwidth areas, older devices, etc.
I've been meeting w/ John Gilmore (EFF) & RMS to make sure we're also testing, to the best of our ability, w/ privacy badger, librejs, and other tools to ensuring beyond nojs support that we have some modicum of assurance for folks who are willing to use js but appreciate a bit more care.
This is a wonderful tool and goes a long way to scratch the "library itch" I've felt since the start of the pandemic. Projects like this make me very happy that I donated to IA this year, and I'd encourage anyone else who likes them to do the same.
One question: are there any plans to sort fiction by author, similar to what is done in real libraries? A lot of fiction is already available in this tool, filed under DDN 8xx, but the groupings are pretty broad.
This is a great idea. One of the reasons Open Library Explorer is possible is we have a healthy amount of Dewey + Library of Congress classification data for our books. As you notes, Dewey has Literature 8xx. Library Explorer also supports Library of Congress (you can go to the settings cog and change the classification system).
Some classifications (LCCN) are better at encodind Author data and we also have a significant amount of author data in Open Library (we'd just need to integrate it more meaningfully in our search index).
Given our small team size, not sure we'll have the bandwidth to prioritize any time soon, but contributions are also welcome from the community and we (the community) meet every Tuesday @ 11:30am to discuss and unblock together.
I'm working on an open source alternative to Algolia called Typesense (https://github.com/typesense/typesense) and just this last weekend I built an Instant Search experience using the 28M Open Library books data set and Typesense:
It never ceases to amaze and dismay me how we've built the greatest library in human history, one with intrinsic properties of unlimited readers per book, negligible distribution cost, and minimal barriers to access, and then done our best to sabotage those properties.
Since the typically imposed access barriers are based on copyright law, the situation could be improved somewhat by:
a) reforming copyright law to allow orphan works to enter the public domain, and
b) enabling authors to (easily) add creative commons licenses (minimally something like CC BY-NC-ND) permitting unlimited readers in digital libraries.
This is visually interesting, but I can't help but feel that doesn't mimic the experience of browsing in a library. It prioritizes book covers that distract the eye (and strain it, being thumbnail-sized). I remember feeling the same way when Apple showed off skeuomorphic bookshelves on the iPad(?).
Are the various scanning projects (inc. The Internet Archive's own) failing to photograph spines? The screenshot of the 3D view in this post shows computer generated spines. Trying things out in the explorer, there's another "3d-spines" view that's similar (and a "spines" view that doesn't seem usable at all, but maybe that's a bug caused by browser differences).
The ideal interface IMO would mimic eyeballing a shelf of spines without the requirement of tilting your head 90 degrees (unless you wanted to, I guess). I'm thinking of a fullscreen interface that resembles the "3d-spines" view, but it lets you drop into a narrow focus on a single shelf, with a stack of proportionally correct scans of real spines that take the full width of the screen, with the hint of the cover in the gaps superimposed by uniformly typeset labels.
I'm sure a simple algorithm could fake book spines based on the dominant colors in the cover. Could maybe even train a ML model if you want to really over-engineer it.
e.g. how the Android app "Blackplayer" colors the music player UI based on the album cover.
But how does one navigate quickly to another Dewey number? I see only next/prior number arrows.
When I use the search box, or when I select Browse > Subjects and click a subject, I am suddenly in a different UI. For example if I choose Browse > Subjects > Chemistry I am looking at a single "shelf" of chem books, and the Chemistry Dewey number and its sub-numbers are no longer to be seen.
"Borrowing" a book just means you can view it on the clunky web reader for one hour? Other online librarians (for the lack of a better word) like Proquest lend books using the Adobe Digital Editions platform, I think that would have been a better approach.
Edit: So it appears OpenLibrary does lend through ADE, but this feature is not applicable for all borrowable books. Some can only be borrowed for one hour, at least in this redesign.
after Borrowing you can call 'https://archive.org/services/loans/loan/?action=media_url&id...' to get ADE to download pdf for 1 hour. Api is hidden from UI after COVID. It was removed to increase availability of books that have only a single copy so that it wont be sitting in someones ADE not getting read. At the same time people waiting in queues to read it.
Not if you have a print-related disability. You can borrow /any/ book published on OpenLibrary.org for up to 14 days. Obviously that is a small subset of the general population (1-2% of the population).
Open Library is an incredible service, with amazing tools. I am sure that this library simulator will be very helpful to people unable to visit a library or unable to read standard print.
I personally have a print-related disability, known as severe convergence insufficiency. Because of this, my neuro-ophthalmologist filled out paperwork certifying that I have a print-related disability, and I have access to the National Library Service, through the US Library of Congress. Open Library gave me reciprocity through that paperwork, and I have access to, and can borrow, every single book published on OpenLibrary.org.
To my knowledge, every preview includes at minimum the basic front-matter (such as those you describe). There are other mechanisms in place which enable specific controlled use cases e.g. limited page previews for folks coming from Wikipedia citations. I don't know if Open Library is presently taking taking advantage of these enriched previews -- I'll check with our product team to see what improvements to the experience may be possible.
38 comments
[ 3.2 ms ] story [ 34.0 ms ] threadThere's something to be said for these physical-world-analogue UIs, in my opinion. Even if this is technically the same affordance as provided by old school web directory UIs, there's more of a sense of tangibility and immediacy to this kind of visual presentation, that makes it more appealing and easier to spend longer times in.
[1] https://openlibrary.org/explore
Maybe I'm just missing it? But the big "Open Library" icon is not clickable, and even the "about" page doesn't seem to have a link to openlibrary.org.
The point is that unless they are extremely well made, the affordance and accessibility is much worse.
OTOH, I lowkey like those environments for social events. Maybe it's just that I'm somewhat introverted, but I find that having a playful/gamelike environment as opposed to a chat window makes interactions less awkward.
FWIW I think the word you're looking for here is "skeuomorph"
Do try out the "Settings" cog at the bottom of the UI -- one thing that makes this interface so powerful is you can add custom queries to transform the entire library (such as to only show kids books, text books, or biographies on any subject).
The way the openlibrary.org site is constructed seems better than the others that came before, e.g., WorldCat. Every page is not just HTML but JSON, too. Omit the slug and add .json after the ID.
https://news.ycombinator.com/item?id=25498127
This is without even considering the availability of bulk data dumps. A+
https://ianews.wordpress.com/2011/03/03/open-library-opds @ https://bookserver.archive.org.
One of the core design features I was aiming for with Library Explorer was that the user should never be navigating the "hierarchy;" they should be navigating the books. Forcing the user to move through categories forces them to try to understand the hierarchy, which can at times be not super user-friendly. Note that the Yahoo/Google examples in the video do just that; the user picks a class, and then sees websites. By showing them the books directly, the user "deduces" the classification tree (lots of books about physics? I must be in the physics section); the classification labels aren't really necessary for exploration.
In general, I think there are trade-offs. A classification tree is a model with flaws (just like all models). But its core benefits are (1) it's a tree; so each node has semantic siblings, parents, and children; (2) each node in the tree has a finite number of child nodes (and usually <30 child nodes; which makes it ideal for human traversal); (3) each book can be uniquely identified by the path from the root. (3) Was very important for physical librarians (classification systems were sort of like a search index for librarians), and was the cause of some of the issues described by Shirky. But online, we don't need (3), so we can kind of throw it away. The tree can be degenerate; books can (and do!) appear in multiple nodes of the classification tree, because, unlike in a physical collection, in a digital collection, books are disjointed from the classification tree itself.
Tag-clouds are _definitely_ more flexible (as Shirky said, classification trees are a _restriction_ on the tag-cloud model), but come at a cost of being harder to navigate (as flexibility usually does) and get a "big picture" idea of. I haven't seen a good example of a UI that lets you navigate tags in a way that doesn't just let you travel from one node to a neighbouring node (please post if folks have one!). Trees allow you to travel up (parent node), down (child node), and left/right (sibling nodes). Graphs you're limited to just traveling to adjacent nodes. You could try to algorithmically _deduce_ parent/child relationships though (that would be an interesting thing to try! I'm sure algorithms exist that do this). Ways of browsing graphs is definitely an interesting problem space! Open Library does also have tags (called "Subjects"), so finding a way to make them more user-browsing friendly would be great :)
My first impression when I looked at the explorer for browsing was "oh, the Dewey Decimal system. I have to figure out which top-level category my interests like, then click on it, and follow the hierarchy of the system down to the books I want. That might not be the intent, but that's what the primary interface is.
I've been meeting w/ John Gilmore (EFF) & RMS to make sure we're also testing, to the best of our ability, w/ privacy badger, librejs, and other tools to ensuring beyond nojs support that we have some modicum of assurance for folks who are willing to use js but appreciate a bit more care.
One question: are there any plans to sort fiction by author, similar to what is done in real libraries? A lot of fiction is already available in this tool, filed under DDN 8xx, but the groupings are pretty broad.
Some classifications (LCCN) are better at encodind Author data and we also have a significant amount of author data in Open Library (we'd just need to integrate it more meaningfully in our search index).
I opened an issue for you here: https://github.com/internetarchive/openlibrary/issues/4319
Given our small team size, not sure we'll have the bandwidth to prioritize any time soon, but contributions are also welcome from the community and we (the community) meet every Tuesday @ 11:30am to discuss and unblock together.
I'm working on an open source alternative to Algolia called Typesense (https://github.com/typesense/typesense) and just this last weekend I built an Instant Search experience using the 28M Open Library books data set and Typesense:
https://books-search.typesense.org/
Since the typically imposed access barriers are based on copyright law, the situation could be improved somewhat by:
a) reforming copyright law to allow orphan works to enter the public domain, and
b) enabling authors to (easily) add creative commons licenses (minimally something like CC BY-NC-ND) permitting unlimited readers in digital libraries.
Are the various scanning projects (inc. The Internet Archive's own) failing to photograph spines? The screenshot of the 3D view in this post shows computer generated spines. Trying things out in the explorer, there's another "3d-spines" view that's similar (and a "spines" view that doesn't seem usable at all, but maybe that's a bug caused by browser differences).
The ideal interface IMO would mimic eyeballing a shelf of spines without the requirement of tilting your head 90 degrees (unless you wanted to, I guess). I'm thinking of a fullscreen interface that resembles the "3d-spines" view, but it lets you drop into a narrow focus on a single shelf, with a stack of proportionally correct scans of real spines that take the full width of the screen, with the hint of the cover in the gaps superimposed by uniformly typeset labels.
e.g. how the Android app "Blackplayer" colors the music player UI based on the album cover.
When I use the search box, or when I select Browse > Subjects and click a subject, I am suddenly in a different UI. For example if I choose Browse > Subjects > Chemistry I am looking at a single "shelf" of chem books, and the Chemistry Dewey number and its sub-numbers are no longer to be seen.
https://github.com/internetarchive/openlibrary/issues/4320
Edit: So it appears OpenLibrary does lend through ADE, but this feature is not applicable for all borrowable books. Some can only be borrowed for one hour, at least in this redesign.
https://help.archive.org/hc/en-us/articles/360016554912-Borr...
There are multiple options for borrowing, depending on the title, including Adobe Digital Editions for applicable titles.
Which can be removed: https://github.com/apprenticeharper/DeDRM_tools
I personally have a print-related disability, known as severe convergence insufficiency. Because of this, my neuro-ophthalmologist filled out paperwork certifying that I have a print-related disability, and I have access to the National Library Service, through the US Library of Congress. Open Library gave me reciprocity through that paperwork, and I have access to, and can borrow, every single book published on OpenLibrary.org.
why not show a few pages from the actual content? isn't that fair use?