the new shading of tags
Join LibraryThing to post.
This topic is currently marked as "dormant"—the last message is more than 90 days old. You can revive it by posting a reply.
not sure this is the right way to describe it. What I mean is the way some tags in the tag cloud have a lighter shade and others a darker shade.
I hadn't given it much consideration till I read someone complaining about it in another thread.
I understood so far that in a tag cloud those tags showing up in a bigger font meant that they were in relation to tags showing up in smaller fonts used more often for that particular work.
However I don't quite understand the new shading.
Could someone explain?
No, the shading is something different from the boldness. I'll look for an explanation of it.
copying my response from the duplicate thread:
I thought the lighter shade ones were indicating it was a more 'personal' tag (example: loaned from Bobby, location: office, etc. etc.) and that darker ones were more 'official' ones (like: horror, fantasy, etc.)
But fiction and non-fiction are also lighter. So it's some other interpretation of what tags are less important.
Can you give an example. Because I look at the latest Harry Potter - and I see tags in a variety of sizes. Some are bold some are not. Although Tim (and others) talk about 'boldness' on the fauxonomy posting it is actually specifying a particular colour for the tag.
Hence tags such as "currently reading" and "fiction" and "hardcover" are lighter because lots of books get tagged with the same tag - so they are not particularly salient. They still show on the main page because they are heavily tagged.
I think there are a few issues around the implementation. For the HP book "read in 2007" is coloured #A3A3FF (less salient) on first looking at the work page, but if you do "show all tags" it is left as the default. I'm not sure why.
In the tag cloud for The Hitchhiker's Guide to the Galaxy, for instance, the tag 42 is light. Based on salience, it should certainly be bold. The lightness is something different--Tim mentioned it somewhere as a new thing, but I can't for the life of me find the thread.
Well the only place I saw mention of the tag 42 is on your thread Option to make all tags dark but it might have appeared elsewhere.
I've just been to The Hitchhiker's Guide to the Galaxy, to have a look and indeed 42 is light, but if you click on it that appears to be correct as the tag 42 is used a lot outside of Hitchhiker's, for quite a broad range of works. If I understand rightly that's salience in action
Are we looking at the same tag page? It occurs 24 times for Hitchhiker's Guide, and no more than once on any given work not by Douglas Adams.
Okay, I finally found the place where Tim mentions it. In this thread, message 12, he says: You'll notice that some tag clouds now show colors. We actually filter the tags for "personal"-ishness. It's not a failsafe thing, though. So we aren't excluding anything—just doing subtle color variations.
Going back to message 2, I find that fauxonomy blog link interesting because by adding in these new shades Tim seems (to my mind) to be using fauxonomy, even though in that blog he seems to be highly against it, quoting it as being "metadata added with the conscious intent to confuse or obfuscate," or to weight them for spammish reasons.
Edited to add: Not that I think he's intending to confuse. But I do think its confusing and unnecessary because it is adding in extra data that isn't user-inputted, by declaring what is 'important' and what isn't.
Hehe. Even if he's not intending to confuse, that certainly seems to be what he's doing--based on this thead, it's obvious that the meaning of the shading isn't at all clear.
It certainly isn't clear. And how is "fiction" a personal tag? I totally don't understand it.
it is adding in extra data that isn't user-inputted, by declaring what is 'important' and what isn't
It isn't adding in extra data, it's just manipulating user data in one extra way. There's no difference in principle between the shading and the different font sizes in clouds.
It certainly isn't clear. And how is "fiction" a personal tag? I totally don't understand it.
I'm pretty sure that fiction and nonfiction "went light" because they have at least some of the same characteristics as more personal tags. E.g., they are applied to an extremely wide variety of books that may be otherwise only slightly related to each other. I don't know exactly what the algorithm is looking for to pick out personal tags, but whatever it's looking for, it's finding in both "unread" and "fiction."
I think it is pretty clear what it means though. Everyone sees right away that lighter means "less important" - that's why people are getting angry about it. "How dare you judge this tag less important, when I think it's very important?!?" Everyone can tell that 42 should be bold for the Hitchhiker's Guide, because we know instinctively that bold means important and light means less so. Just because we don't know the exact algorithm being used doesn't make the effect unclear.
well thanks all for your intents to enlighten me on that subject :-)
However I keep not knowing what the rule is for darker or lighter shading. On what base exactly does it happen. it all seems rather arbitrary. Kind of done manually on a case to case basis and thus giving a feeling of being based on personal feelings and judgement of whoever does the shading.
the overall effect is confusing
In the end I keep thinking that it wasn't a good idea. For my part I prefer the old traditional tag-cloud system.
actually someone wrote about an option to enable or disable this shading. That would be a nice feature.
Along with "view numbers" and "show all" you could add "disable shading" on the social info page.
Speaking of which when you click "view numbers" it might be a good idea for the "link" to change to "hide numbers" which it doesn't do in the actual state.
And then I don't quiet understand why there isn't a "feature page" or something similar where the LT-team explains the new features it has added to LT. Would serve the community and save us a lot of confusion. Instead of having each time to post a new thread asking for explanation and more often than not getting a whole lot of posts of other community members doing guesswork about the new feature but no clear and concise answer from the developers.
so far as a workaround there is Stalking Tim which aims to track where he and his co-workers do post about their new features.
It would be good if they were centrally published though. It would also be good if there was an ideas board where we could see what was on their radar to be implimented "soon", "some time away" and "only in a month of sundays".
I'm not too keen on the shading either. I'm perfectly capable of deciding for myself if today the tag "unread" is important to me, and tomorrow if I'm more looking for a genre tag. My need for others tags changes. The shading doesn't.
Yet another distraction from the real job of enhancing the cataloguing interface. I can't see that it adds any value.
I actually prefer that things are sprung upon us: it makes it interesting to go around the site every now and again to see if there's anything new to discover. I would really like to see what's on the to-be-developed list, and some indication of priorities.
well some us do have a job and even family and would prefer not to have to spend their time browsing/going around the site in search of explanations about new and sometimes confusing features when it would be easier for everyone if those explanations where centralized on a page and posted at the same time as the new features are added instead of having to wait for enough members to make enough fuss and noise to trigger an explanation.
There's no difference in principle between the shading and the different font sizes in clouds.
I think there is a difference. The boldness and size tell us something about the numbers of tags; large means "a lot of tags" and bold means "a lot of tags compared to the overall usage of this tag". I don't know what algorithm Tim is using for the shading (obviously, since he hasn't bothered to explain it), but it certainly gives the impression that it's not purely about the numbers. And the fact that it doesn't convey any clear information is what makes it irritating and almost offensive to me. It's not telling me anything about the tags; Tim is just telling me what tags are less worth looking at than others, in his opinion. And frankly, I don't want to hear that. As reading_fox said, I can make that decision for myself. Unless the shading conveys new information, it's just obscuring the real data and imposing Tim's views about the worth of various tags on us.
I hate the shading. I have enough problems with my weak eyes, and now this . . . mess.
Tim seems to be rather fanatical about our freedom, so I don't think he's trying to impose his views on us, but is trying to tell us something interesting or useful - it's just that we don't know what it is.
#23 maybe we should open a petition thread were everyone could sign for Tim to explain things ;-)
#22 I totally agree, the tags with lighter shading are sometimes difficult to read
#21 well the shading does convey new information: it tells you which tags are more or less important in Tim's view ;-)
What makes you think this has anything to do with "Tim's view"? I seriously doubt that Tim has gone through all the tags and marked them as more important and less important. I think he has some algorithm based on the data in the database, its just that we don't know what that is.
Well, maybe it's not an intentional imposition, but a thoughtless assumption that we value tags in the same way that he does. Either way, the effect is the same.
#25--The post of Tim's that I quoted in message #11 seems to make it pretty clear. He's making the personal tags lighter. This means that in his view, the personal tags are less important. The problem is that not everyone feels that way.
It isn't clear from what he said in that thread whether it was an algorithmic filtering or a list of tags to filter against. As new personal tags are likely to appear on a pretty regular basis I would imagine that it is algorithmic in nature.
However it is futile for either you or me to second-guess the system.
#25 it doesn't really matter whether he does it manually or with an algorithm. To use an algorithm is to just to automatize what you would do manually. An algorithm always reflects it's creator's opinions.
Does I get it right, in a tag cloud we have three elements:
1. Size of the Font
2. "Boldness" of the Font
2. Shading of the Font
Please can some developer explain us the RULES for these three elements so that we can stop the guesswork?
And not everyone feels that they are as important. Not everyone feels that they should be shown at all. Not everyone thinks there should be a Talk section of the site. Not everyone thinks tag combination should be allowed. Not everyone even thinks work combination should be allowed. Not everyone thinks reviews copied and pasted from somewhere else should be flagged. Not everyone thinks "hate speech" should be allowed in Talk posts. How many of these problems of disagreement concern you?
I just don't think any of us can talk about this productively or substantively without knowing the algorithm that is used. Maybe the problem is that people keep referring to them as "personal tags" being lighter, but then say things like "but 'fiction' isn't a personal tag." Okay, then personal tags aren't what are made lighter. What are made lighter are tags that fit filter X. So "Tim's view" isn't that personal tags are less important, it's that X-type tags are less important. And I have a feeling that logically, if you saw the algorithm, you might say "yeah, that does pick out things with less salience." Just because sometimes you're interested in those things, or even if you were interested in them all the time, doesn't make them salient logically speaking.
"So "Tim's view" isn't that personal tags are less important, it's that X-type tags are less important."
You can't even really say that, just that Tim wants to represent it some way in the tag cloud and shading is one way to do that. Did he say anywhere that he intended light shades to mean less important?
Okay, Tim's view is that X-type tags are less important. A large part of the problem is that this new "feature" was introduced without any explanation, so instead of gaining any data, we're just supposed to accept that X-type tags really are less important.
I don't think it's about salience, though, at least not in a purely numerical way. Whatever algorithm determines boldness was working fine for salience. This new one makes errors like making 42 light. Also, in Nickel and Dimed: (colon added to find touchstone) the tags class and work are light. It seems like this is on the assumption that the book is used for class and work, when actually it's about those things.
I usually try to suggest compromises that will make everyone happy; I posted a couple of weeks ago asking for the option to turn the shading off. This is the sort of issue that doesn't have to be decided one way or the other. We could decide what worked best for ourselves if only they'd let us.
There is no algorithm that is going to tell the difference between the tag class meaning that a book is used for a class or is about class.
This is why the algorithms should be based on numbers and not on assumed meaning.
_Zoe_, I would agree that class and work are not correct, but these issues smack of a work in progress.
Give Tim some time and see if these problems are corrected.
There is no algorithm that is going to tell the difference between the tag class meaning that a book is used for a class or is about class.
For one particular user you are correct. For the community, a tag's related tags may give an indication.
The related tags for class see to indicate that most folks use the tag as meaning about class.
The related tags for class and the tags used on Nickel and Dimed: do seem to intersect well.
This begs for deep analysis on the tag cloud/tree. Doable, but expensive in computing resources. I am just speculating, but this may be why Tim is going slow.
still, some tags are shaded in such a light way it get's difficult to read them.
Why not give people the option to choose to see a standard tag-cloud, on with the numbers, one with shading, etc. This way everyone get's what he wants everyone is happy.
37: pivox, I don't believe anyone disagrees, but its unclear if Tim or any developers have even taken notice of this thread.
33: Zoe, I think the the algorithm *is* based on numbers (well, numbers and relationships in the database), what else could it be based on?
Tim did read the other thread with the suggestion to make it optional, but didn't say anything about the suggestion itself.
It could be based on both numbers and an arbitrary list of words that Tim deems unimportant. If it is just numbers, it seems to be making decisions based on how the tag is used throughout the site and disregarding whether it actually is important for a particular work. Looking at the search results for the tag school, for example, school shows up light in the tag cloud for all of the top ten results.
Looking at some of the shaded tags on H2G2 it seems that the shading is specified by an explicit style attached to the anchor. It should be pretty simple for someone to write a greasemonkey script to remove the shading - although of course that will only work for Firefox. If the markup was changed so that the style was picked up through the use of a css class then it might be possible to have two different css files - one with shading and one without.
I have no idea how to do that, but looking at the tags does answer jjwilson's question in #30: Yes, he says explicitly that the light tags are less important.
Hi there. I'm sorry I only just saw this thread. To be frank, the FAQ group isn't in my (otherwise rather exhaustive) group list. I have so much ground to cover that I tend to stick with things I've posted to and the groups I'm in. I've gone ahead and started watching this one. I know I'm sometimes hard to reach, but whenever a long discussion of my intent goes on, I hope I can drop into it!
The tag color was an experiment. Frankly, until seeing this, I thought it was largely unnoticed. Actually, coloration had been going on for a while, but only on *new* tag clouds. (Large tag clouds can take a while to calculate; a tag cloud becomes new when the underlying number of tags changes by a certain percentage. Small clouds never get cached; when you ask for "all" tags, you get a fresh one no matter what.) Anyway, we changed to a new caching system recently and that made everything new, and the colors suddenly pop-out.
Coloration was based on a judgment of whether a tag was personal that was partially algorithmic and partially not.
The "not" part was actually hand-done. A key selling point of LibraryThing for Libraries is that a real live librarian ("REAL LIVE LIBRARIAN!") went through the LibraryThing tags and filtered out the personal stuff. Selling tags to libraries is hard enough, but if there's one thing that libraries don't want it's books tagged "at the beach house." So, we made a cut. We only considered tags used 15 times by 10 users (or some such number). That came to 25,000. Abby and an intern, Jenny, both MLS-carring librarians, then went through them by hand. (Then they both dropped dead. We're hushing it up.) By default, low-incidence tags were marked as personal.
I put the colors up as a test. I wasn't sure what I thought of them, but the idea was to see what others thought, and then to expose a system for users to mark whether (or perhaps to what degree) a tag was "personal," starting with the data we'd accumulated. Frankly, I saw no purpose to introducing that system until people had signed off on the colors. Without the colors, it didn't matter if some tags were personal and some weren't. I do apologize for not telegraphing what I was doing but, frankly, I wanted to see people notice it and start a conversation, not make some big blog post about it and later decide to chuck it.
Since the topic was mentioned, I *do* think there are better and worse tags, and that it might be useful to distinguish personal ones. What those are is largely dependent on what you're trying to do, but when it comes to other people's tags, "subject" tags are better than most personal ones. Now and then there's a funny exception--I love the "unread" tag, for example. But I don't get much out of browsing the birthday tag, or some of the unique numerical systems people have developed. I certainly don't think those tags are bad to enter—I enter many. But there's an argument for making them less prominent in clouds, I think.
Reading this, I *think* the experiment is a failure. People don't like the colors. They find them hard to read. They are not explained. People are suspicious of the motives underneath it. (And if you don't like my motives, I'm guessing you won't like the mob voting on the personal-ness of your tags.*)
So, I'm inclined to drop them. Anyone want to stop me?
Some quick responses:
nperrin: "It isn't adding in extra data, it's just manipulating user data in one extra way. There's no difference in principle between the shading and the different font sizes in clouds."
I disagree. Look, tag clouds are a *representation* of the data. They aren't the data. That's why LibraryThing (alone, as far as I know) has a link to show you the actual numbers involved. We care about the data underneath.
The simplest tag cloud is one in which the font size is directly proportional to the number of tags applied. But if LibraryThing didn't adjust the sizes in various ways, all sorts of tag clouds would be ten items in 7-point type and one in 500-point type. In the author cloud, J. R. Rowling would be four feet tall. In fact, we apply progressive throttling so that we end up with readable blobs that also show some of the peaks and troughs of the data. The result is a manipulation in the way that all simplifications are a manipulation.
*Note: Few seem to object to tag combination, which also reduces freedom to some extent.
*dances a bit* Yay, a response from Tim!!
I still don't like/agree with the new shading, especially since it seems so random (I know it's not, but so many tags that are certainly not personal are being lightly shaded...). However, I am at least a little more understanding/relieved now that I know what's going on and what Tim has to say about it.
I think that the main problem was the "they are not explained".
Personally I don't mind the shading. If I did I would have taken action (via Greasemonkey) to make them all the one colour.
I'm going to write a Greasemonkey that changes your "colour" to the correct spelling ;)
I like the shading, and if it's leading to a place where users will be able to affect it I think that's great.
The explanation does make it sounds a bit better. But I'd still say no overall.
The first problem is that the lighter colours are annoying to read. I might be willing to put up with that if they were telling me new information, but it's not worth it if the information is just the result of a poll about what tags people think are important (I think it's better to talk in terms of important rather than personal, because some tags like "unread" are personal but still important, while non-personal ones like "fiction" have been deemed unimportant).
I think a second problem is that what works for LT isn't necessarily the same as what works for libraries. The libraries might want to exclude any potentially bad or useless data from their tags, but seeing everyone's quirky tags is one of the interesting parts of LT. I'm not interested in a cleaned-up version of the data (especially because the truly useless data isn't very prominent anyway); I want to see what people really say about their books.
Lighter tags that I like (and want to be dark) include fiction, non-fiction, read, unread, TBR, favorites, series (it's very important when coming across a book for the first time to make sure it's not the fifth in a series), lt-inspired, etc. Even tags like borrowed, owned, and wishlist can give you an idea of how much people like a book or author.
Then there are the light tags that really should be dark, given their usage: 42, class, work, school, pi, etc.
Tags like birthday are pretty useless, but they aren't prominent in tag clouds anyway, so I don't think it's worth the effort of getting rid of them.
My first instinct was to say go ahead with it but give us the option of turning it off. If it's going to take to much development time, though, I'd scrap the whole thing.
Since two mention liking the shading, or not minding the shading, I'll just again note that I can't read the light shading. Just didn't want to leave it seeming as if everyone now agreed.
My main problem was with the shading :)
edit: dang, this was supposed to before Zoe's :) and after the two that mentioned liking or not minding the shading :)
I like the idea of the shading - now that I know what it means. Maybe designate personal tags in a different fashion?
Have never liked the shading. Can't read the lighter shades and I don't care for the judgement call about what tags should be more prominent.
#42 when you ask for "all" tags, you get a fresh one no matter what
Is there also a way to refresh tag mashes and tag pages? See http://www.librarything.nl/talktopic.php?topic=18509
I agree with Zoe (Message 47:)
I'm no fan of shading, not even after Tim's explanations (worst thing is the "difficult to read" aspect).
I must say I wouldn't mind the shading if it were optional as is currently the option to see the numbers. I think that would be the best solution, to let members choose (in this case and I would say in most of the cases since LT has grown to be such a big community and hence necessarily a very varied community with different needs).
And I keep thinking a better way to approach it (even though it could be complementary with the shading on the condition that shading would be an option) would be to give members to possibility to make some of their tags private. Obviously not everyone would use this possibility but it would certainly help.
While I do like shading (The concept behind it. The "difficult to read" is just a user interface design issue and can be addressed), pivox (54:) has good suggestions.
A question for Tim. If shading goes, what thoughts to do have on other approaches?
I like the shading. It's fun to read the light-colored ones to see what kinds of obscure and strange tags people use! :-)
This was a hard one, but I've decided to get rid of it. They're cached, so they may take a while to fade out.
Thanks, Tim -- I definitely get the libraries issue, but readability trumps that for me.
IMHO, you're trying to address this issue at the wrong point in the LT workflow.
I already know which of my tags are "personal" (e.g. "BOX5" or "toread") and which aren't (e.g. "economics" or "math")--though admittedly there may be a couple that have the potential to go either way depending on whether other folks adopt them (e.g. "blewmymind"). If I had a way to tell LT which tags are personal up front, you could filter them out automatically, and improve results while simultaneously reducing processing.
My simple proposal is to treat tags starting with the underscore character as "personal, i.e. not to be aggregated. (This would be in the same spirit as the "rel=nofollow" convention for anchor tags in HTML.) Of course, any character would work. I like the leading underscore character because it's often used to denote private objects in OOP and because it's not heavily used for something else (e.g. for regexes).
True, existing tags (including mine) would have to be renamed by their users. But your tag renaming facility makes this easy to do in bulk, and the heaviest users of personal tags are probably also the most active and geeky LT users, so I suspect that it would work itself out pretty quickly.
BTW, you can also return value to me. Right now, I (like some others on the site) catalog my "personal" tags in the description of our profile. You could do this for me automagically. You could also offer the option to omit personal tags from personal tag clouds, etc.
This takes care of "personal" data. The other issue is malicious data. I believe you could address that in an automated fashion, as well. If someone creates a profile in order to spam ratings for certain authors, or just generally fsck with your site, that profile ought to look significantly different to your ranking and tag processing algorithms than other profiles on the site. It seems there must be a way to identify such "deviant" profile automagically.
Again, that won't address the issue of someone who generally tags and rates responsibly, but intentionally screws with the data for a particular book, tag, etc. But, then, that should by definition be a tiny percentage of the activity on the site.
BTW, I do understand that having a Real Live Librarian currying data on the site is a selling point for your LT for Libraries program. But, just as Google uses its humans to tweak its superior automated search results, IMHO you'd do better to have LT's humans add value elsewhere in the chain.
>But your tag renaming facility makes this easy to do in bulk
Only if there's a way to preserve the order... and some people use other special characters to sort within those categories.
That's one of those seemingly simple features that never gets implemented. There is in fact no Replace function in power edit, unbelievable as that sounds.
> 60 But your tag renaming facility makes this easy to do in bulk
I did that more than a month ago, but the tag pages and tag mash pages don't get updated. All the books I don't share with anyone prove that, they still appear on the wrong/old tag pages and tag mash pages.
Ugh. I forgot that tag renaming is still a two-step process. So, yeah, it would be good to fix that first.
I'm curious about the significance of tag order, though. Does this affect something I'm unaware of?
Tag order - some people sort by the tag field and order helps somehow or another here. I have comments and tags combined into one column, so I've never been able to do it. So I cannot say exactly the significance of tag order.
If you have a lot of tags it helps to keep them in order by the type of tag so you can see if you're missing one. If you have hierachical tags like non-fiction, science, biology, neuroscience, then the order makes even more sense. Or if you put a location tag first you could use it to sort your books by location. There are lots of reasons to keep your tags in a certain order.
MikeBriggs: then switch your column order to separate tags and comments ;)
Ah, but I enjoy having the extra column that combining tags and comments gives me. :)
However, combining tags and comments in one field-column doesn't allow you to edit that field :-)
I do not understand. I can still use power-edit to edit that field. Oh right - sorry, I never change anything by editing the individual field, and when I want to change something, I go to the individual work or power-edit. I forgot the "editing" field option.
Thanks for reminding me. That is a much quicker way to fix my book's titles :)
I'm with you there. On another thread, I supported someone's suggestion to let us have more columns: until that happens, tags and comments combined will have to do.
73> I just went through all of my view selections - There is just so many columns that I want to be able to see, but I note that one of them does have tags and comments separate. Just one of them though.
Again, thanks for reminding me :) I've been double-clicking and changing tags in that one view where I have them separate :)
You are right of course, I was talking about the view I use most of the time. I do use other views for special purposes. But I would still like more columns available. I appreciate the problems of people with 800x600 resolution used as 600x800, but for those of us lucky enough to have 1920x1200 available, more columns would be nice.
76 > nor was I trying to stop the request for more columns, I'm sure if they were available I would expand to use them as well.
But until such time I try to optimize different views for specific groups of information and try to remember which is which.
#66 - tag order - its the only way currently to get a series to appear in order in your catalog.
The series designator tag has to be the first tag in your tag field for this to work. Try looking in my catalog for the author C J Cherryh. Sort by author - the several series are an intermingled mess, if you are unfamiliar with her work you won't even know which belong in the same series. Now try sorting by tag, much much prettier.
This topic is not marked as primarily about any work, author or other topic.