Make separating a simple majority

Right now, both combining and separating tags takes a super majority. Unfortunately, there are a large number of tags that were combined in the old days in ways that problematic now. For example:

http://www.librarything.com/tag/Loeb+Library
http://www.librarything.com/tag/Loeb (includes "loeb, loeb classical library, loeb classics")

The combination of Loeb and Loeb Library is getting voted down because Loeb is also a name. However, the separation of Loeb Classical Library from Loeb won't go through, because some people look at Loeb and think it should be combined with Loeb Library and friends. What it's leaving us with in this case, as well as many others, is the worst of all possible worlds; we don't have a Loeb tag separate from the tags that are explicit about referring to the Loeb Classical Library, but Loeb Library doesn't get combined with other tags referring to the Loeb Classical Library.

To solve this, we can make separating a simple majority. This will stop a few people from stopping the separation of Loeb and similar tags from more complex tags and permit the tags to be recombined in a way that makes sense. There's generally more support for separation then combination, so making combination a simple majority wouldn't usually fix the problem; making them both would work, but would completely override the protection that making it a supermajority was supposed to give.

Yes, please. It's never made sense to me that combining and separating both require supermajorities. If tags are only to be combined when there's a clear consensus, then it shouldn't be so difficult to undo an existing combination for which support is lacking.

This is an excellent proposition from prosfilaes, and I give my support. I don't know if it has to be a simple majority, but the threshold should definitely be lowered from what it currently is. These days, simply having 3 or 4 "no"-votes on any combination proposal is enough to essentially kill it, i.e. keeping the tags separate.

So why should active separation efforts be made to jump through way higher hoops? Especially when the tag set is full of these bad historical combinations that really do hamper current good combination efforts.

With easy ability to kill any proposal with just a few "no"-votes and with this separation bug, I am surprised that anything ever gets separated these days.

I have never 'combined/separated' a tag. Though I do understand why it is important to some.

BUT... perhaps, if the defenders of "NON Separation/Combination" were put on notice, and required to argue their case (within a reasonable time)

The default might/should be a simple majority?

I suppose those who have a 'strong' point to make, will be heard. The others, who probably only made an error or who are conservative by nature can be ignored.

In case I inadvertently gave the wrong idea in my last post, I don't really have a problem with a rather small minority killing combination proposals. I think this is usually healthy, and have had a number of my own proposals killed this way, only to realize much later why people probably voted against them. I just find it odd that these are effectively separations achieved by a small minority, whereas achieving actively proposed separations requires super-majorities.

I think the various "Why I voted No/Yes/Undecided" threads in the Combiners! group do a relatively decent job now of what guido47 discusses, though they are voluntary of course.

Another solution to try if TPtB don't like this one is to separate voting for combinations from voting for separations. I can't tell you how many times I've caught myself voting "no" on a separation because I thought it was a combination. (I know, I know, they're pink, not white. But I get on a roll scanning the right hand side of the column and the color just isn't enough sometimes to make me stop and think.)

>6 aulsmith:
Where are you voting from, aulsmith? I haven't yet seen anything that's pink vs. white. On tags_combinations.php, I see gray for combos and yellow for separations (with confusing numbers prefixing the separation links). On the helpers log I see no color distinction, and separations are called combinations (bug reported). And on the tag page themselves (see example), I see everything in the green block at right, all under a misleading heading of "Proposed combinations", where each proposal inconspicuously notes whether it's really a combination or separation.

I'm worried that yet another place exists that I don't know about where it's confusing what's going on and causing people to vote the opposite of what they're intending.

I think the color difference may just be a browser difference. I'm using tag_combinations.php.

5> That's just because you're a splitter; you're more concerned about properly identifying categories than whether a particular combination might be useful despite it being inexact.

If you were a lumper than you would say that separation failures are effectively combination achieved by a small minority and complain about how unfair that is.

ETA: Or maybe you'd realize that the failure of either a combination or separation proposal is just that and isn't effectively anything else.

> 9
I don't know when I've ever self-identified as a "splitter", and I haven't complained about things being "unfair" to me. I have generally tried to follow the combining guidelines which seems to caution against lumping, as well as the system-stated "rule" (see green box at right) which twice states "when in doubt, separate it", and in the same place claims that the "system is designed so it's easier to separate than combine". But maybe the mess I made WRT to the London tag made you assume I have some agenda to go splitting everything up into precise little categories. I don't know.

What I have expressed frustration over is the impasse created by the fact that pre-vote combinations and perhaps other factors have created many situations in which the winningest alias becomes a gatekeeper that lets nothing in and nothing out. This is equally "unfair"/"fair" to both lumping and separating. Lowering the separation threshold would I guess seem "unfair" to lumpers, but it at least might create some movement, presumably in the direction favorable to TPTB (I could be wrong, of course).

> If you were a lumper than you would say that separation failures are effectively combination achieved by a small minority and complain about how unfair that is.
(Edited) This kind of argument doesn't really hold much weight for me when taking into consideration all the pre-vote combinations. The argument would make more sense if voting had started with all tags separated. They enjoy special status because they never had to pass any threshold.

Maybe this whole thing would be moot if tag disambiguations (and/or bisection and/or moving) were implemented, but those also presumably require much more coding than prosfilaes's recommendation.

This kind of argument doesn't really hold much weight for me when taking into consideration all the pre-vote combinations.

You're forgetting that pre-vote separations also did not have to pass any thresholds, so there is and never was any bias in the system towards either combinations or separations.

I have generally tried to follow the combining guidelines which seems to caution against lumping, as well as the system-stated "rule" (see green box at right) which twice states "when in doubt, separate it", and in the same place claims that the "system is designed so it's easier to separate than combine".

Well, the tagging guidelines are written by users and with the majority of splitters over lumpers I'm not surprised that they're written that way. But I never read that green box text before. I wonder how long its been there or when it changed. In Talk Tim has always seemed neutral on the lumper vs. splitter wars but that system text is pretty explicit.

But then again, Tim did think that humor should be separated from humour but he lost that one when he gave users a vote in the matter.

In any case, I disagree with the official text and the guidelines and Tim gave me a vote so I'll exercise it as makes sense to me. And I disagree with any proposal that will diminish my vote.

9>

Having both situations require a supermajority, when in the past combining could be done by a single user, has produced untenable results.

There are numerous cases where some variants of a plural form are mixed in with the singular, but others aren't. Attempts to combine the plural and singular fail; so do attempts to separate the plural from the singular so that it can be all plural forms together and all singular forms together. The status quo is not what anyone would prefer, but since neither the "combine singular and plural" nor the "separate singular and plural" side has a large enough majority, it's what we're stuck with.

I'd support making separation easier than combination for tags because (1) the type of problems that exist are more easily solved by separation and (2) there are many fewer separation proposals than combination proposals. In my case this is a pragmatic approach.

#12: And I disagree with any proposal that will diminish my vote.

Funny, I thought the goal was to get a good tag system, not to cultivate personal power. I will say I can't recall a smaller power grab so aggressively stated as such.

14> Oh please! Isn't the power grab those who want to make it easier to separate than to combine? In any case, I believe that in many cases combining is better than separating, even if they are technically different, such as singular vs. plural and my opinion is as good as anyone else's.

11 > [T]here is and never was any bias in the system towards either combinations or separations.
To me, the system is very biased toward separation by virtue of the fact that all tags with the same orthography (excepting case-insenstivity) start out separate from one another. Imagine a parallel universe where all tags start out in a giant tag-lumping pot from which they had to be separated and regrouped by voters. And then tell me that wouldn't be biased toward combination. Would separation even be necessary as a feature these days if it weren't for the bad pre-voting combinations? (In fairness, the other reasons I can think of are (1) lack of pre-threshold-meeting vigilance (many proposals meet threshold in less than 24 hours) and (2) more general changes in the combination hive mind over time. And reason 1 strikes me as another good reason to lower the separation threshold in our current a-few-no-votes-kills-any-combination world.)

12 > I wonder how long [the green box text has] been there or when it changed.
If the archive.org snapshot is to be trusted, it's been there since at least 2007/10/30. Oh, I also see a mention in a discussion thread dated 2007/05/22.

The problem is that we have two often opposing approaches to tag combination among LT users.

1. tag combination creates connections between books (and users) by combining tags where the user tagged the book with the same intention. So if I tag a book "London", I want to find all the books that have been tagged with the same intention.

2. tag combination is an exercise in data quality. Tags should be combined or separated based on theoretical differences in meaning and usage. If someone tags a book "London" and someone else tags a book "London UK", the tags should be separated because "London" is ambiguous.

While there are these two approaches, the super majority rule will result in contradictions.

The solution is not implementing an absolute majority. In fact, that is the worst possible solution. A majority is not a consensus.

The solution is a system where a Yes vote promotes the connection between two tags and a No vote demotes it.

The same can be said of image voting on LT. We have a group of about three users that probably account for 90% of image flagging. While LT's manifesto is to strive for greater independence by creating connections with more authors, it seems like we have an author per week totally pissed off with LT because the image they use to represent themselves on Goodreads, Amazon, Facebook and/or their blog is not good enough for LT.

It is the worst thing both from a statistical point of view and a social psychology point of view to have this absolute voting in several places on LT. Voting to achieve consensus should only ever promote or demote. Not destroy.

>17 marq:
OK. But what would that mean in practice? In image flagging, either the image is displayed or it isn't. What would it mean to demote an image without removing it? In tag combining, either the system treats the tags as the same or different. What would it mean to promote two tags as similar without making them the same?

When an image is demoted, it is not the primary image. The most promoted image is the default or primary image, other images you have to do something to see them.

With tag combinations, there is already something similar in tag mashes where you can use an operator to demote. A user sees the books tagged with the primary tag - the tag they are searching for, followed by the most highly promoted related tags. There has to be an algorithm that takes into account the number of books tagged etc..

PS: Keeping in mind though that most tag votes seem to be 100% Yes, so they would have the same weight as the primary tag i.e. work in the same way as fully combined tags do.

> 17: it seems like we have an author per week totally pissed off with LT because the image they use to represent themselves on Goodreads, Amazon, Facebook and/or their blog is not good enough for LT.

What puzzles me is that such authors see author pages, which merely are about them, as their own personal playground. They can have whatever pictures they like on their profiles.

#19: With tag combinations, there is already something similar in tag mashes where you can use an operator to demote.

Having some idea how this is implemented behind the scenes, let me say not a chance. If I'm not mistaken, right now merely asking the question of what forms of tags are used for works would be a very expensive operation. Anything like what you suggest would have complex hard-to-work out semantics* and be very, very slow relative to what exists.

Authors are attempting to connect to their readers through social networking sites like LT. These days, anyone can be an author. You don't need to convince a publisher of the merits of your work, you just need to put your eBook up on Amazon. Many of these authors, for a wide variety of reasons, wish to remain anonymous and write under a pseudonym. They use a particular image to represent themselves which for obvious reasons may not be an actual picture of the author. These kinds of pictures uploaded by authors are routinely kicked off LT, although often used by the author on GR, their blog, Facebook etc..

Regardless of what we think about the merits of the picture an author has chosen to represent themselves with, that is the image they have chosen. If they can't connect to their readers using the image they chose, they simply wont bother using LT.

They can do as they please on their profile pages. Author pages are more like Wiki than like Facebook. They won't be able to control what's happening there anyway.

Well we can argue this point endlessly, an author's persona as a pseudonym is different to the author as an LT member with a profile. This thread is about voting. We can have different views about what images can represent authors. The problem is that there is no space for consensus. It is win or loose. If you loose, your picture is deleted. If your picture is deleted, you are not happy about it and you draw certain conclusions about the culture of LT based on the opinions of a very few people.

Is authors connecting to their readers via LT something LT wants to encourage in the first place? The "LT for authors" page or whatever it's called doesn't exactly radiate welcomingness to the would be connectifying author.

This thread is about voting.

Fair enough. I was indeed digressing.

25: You may like to read http://www.librarything.com/topic/44126, second last paragraph from the end of the first post:

"Independence. We're (LT) already pretty independent, but we can, I think, continue to exploit that. We need to hook up with more independent bookstores. We need to hook up with more authors, publishers and libraries--anyone who will make the site more interesting and no one who will bend the site toward monotony (sic) (an) intellectual monoculture. "

Well, the whole article is interesting. But really, if LT is trying to avoid an intellectual monoculture, it obviously needs to get rid of mechanisms that are anti-consensus.

17: I think your explanation in 2 seriously misrepresents the "splitter" position. Folksonomies, which is what LT's tag system was supposed to be, were supposed to allow users to make very specific and personal divisions of their material, which was impossible with controlled-vocabulary subject systems that are the norm in libraries and academic settings.

So, for instance, if I want to use the term "Ray-gun" to describe the 40th President of the United States, I'm free to do that. If you folks come along and combine that term with words denoting science fictional weapons, you've destroyed my work and any meaningful connections that I might have gained from it. Splitters are really conscious of this and want to avoid that kind of problem.

One could, of course, argue that it's impossible to tell most of the time what the heck anyone meant by any specific tag they used and that we'd be better off with a controlled vocabulary system. I would be in favor of that.

But until LT management has said that they give up on folksonomies and they want books in standardized categories, I'm going to continue to help them work towards a folksonomy. And making it easier to split words that have been inattentively combined would be a big help in that direction.

29> So, for instance, if I want to use the term "Ray-gun" to describe the 40th President of the United States, I'm free to do that. If you folks come along and combine that term with words denoting science fictional weapons, you've destroyed my work and any meaningful connections that I might have gained from it.

That's completely false. Combining tags only affects the social use of tags. It has no effect whatsoever on how the tags are used in your catalog. You cold still search to Ray-gun within your catalog and see it on your tag page and it will still be attached to the same books.

But when you get beyond your catalog to merge information from many people then trying to preseve such idiosyncratic meanings of tags no longer works. What's the use to the general community of knowing that one person uses Ray-gun to refer to Ronald Reagan? Whose going to remember that if I use the hyphen that I'm referring to the president but if I don't then it's weaponry from science fiction?

30: But when you get beyond your catalog to merge information from many people then trying to preseve such idiosyncratic meanings of tags no longer works

I don't necessarily disagree with you, but there are many people who are doing a lot of thinking about subject retrieval who do. LT is an experiment to see if it can work. By your voting, you're saying it doesn't work. However, I think LT's management wants it to work, which is why the guidelines are written the way they are. If, indeed, they still think it will work, implementlng this policy would be helpful for the reasons given in the original post. I'm still willing to go along with the experiment if they still care about it.

But using London to mean London, UK is folksonomy in action. To refuse to combine them is to defeat that purpose in favor of some stale exercise in pure classification.

>29 aulsmith:
Thing is, it's my memory that the "Raygun" nickname for the president was precisely a reference to science fiction weapons based on Reagan's support for the "Star Wars" missile defense program. That this one tagger used "ray-gun" instead of "raygun" or "ray gun" is of no particular significance. If you want to capture the folk meanings of raygun then Ronald Reagan is among them.

28 > Thanks, interesting read.

I do confess, to my ears "Authors are attempting to connect to their readers" sounds suspiciously like "spam". But that's got precious little to do with tag separation.

#30: But when you get beyond your catalog to merge information from many people then trying to preseve such idiosyncratic meanings of tags no longer works.

It's not particularly more idiosyncratic then many other uses. I'm the only user of a number of tags, and if you threaten to combine them with different but less idiosyncratic uses or pollute them with different uses for the lulz, I will be deeply annoyed.

Yes, Ronald Raygun or Ronald Ray-gun works better then Ronald Ray Gun, whereas raygun is rare for the name of the item and ray-gun just looks silly. It is of some significance that ray-gun was used for the president instead of any other form.

#32: Were it that simple, there would be no argument. But at least three separate users on three separate books--On The Beat: 150 Years of Policing in London, Ontario, London street names*, and Sketching rambles in Ontario--use London to mean London, Ontario. Now you're talking arbitrary thresholds, not facts.

* I had to combine this, so a naked eye inspection no longer shows that three different users use London for London, Ontario, but here's direct links to their catalogs.

http://www.librarything.com/catalog/douggeo&tag=london
http://www.librarything.com/catalog/joangerber&tag=london
http://www.librarything.com/catalog/Deejay&tag=London

30, 32: We have obviously read entirely different materials on how folksonomies are supposed to work My understanding was that it was just those idiosyncratic, sub-cultural references it was supposed to be good at preserving.

London vs. London UK is a perfect example of why I think this whole project is doomed. Clearly everyone using the tag "London" knows what they mean, but the only way to separate actual meaning is to get them to recognize that other people use the same word to mean something different and force them to use a controlled vocabulary (Which is why if I ask weather.gov for weather in Albany it asks me if I mean NY, OR, CA or GA.) The only harm in combining "London" with "London UK" is that you take a meaningful tag "London UK" and add noise to it. Depending on how much noise gets generated, it can eventually make the tag meaningless. I'd rather leave "London" (which is meaningless now) alone and keep meaningful tags like "London UK" meaningful.

33: Again the problem is signal-to-noise. If the few books labeled Ray-gun and about Reagan get drowned in pages of raygun sf fiction, you've lost the point of using Ray-gun as a tag in the first place.

30>

Combining tags only affects the social use of tags.

That is the significant point on which I strongly agree with you. Tag combination is purely a social function of LT. However it is also the point on which people will scream their disagreement.

We combine "London" with "London, UK" because we suppose and see that the vast majority of people who tagged their book "London", meant "London UK" (and vice versa). We create the connections between those books tagged "London" and those books tagged "London UK" (but not both).

In the process, we create a few inappropriate connections between books tagged "London" where the tagger meant "London, Ontario" with books tagged where the user meant "London UK". This is mitigated by the fact that we have done a lot more good than harm and also that anyone who tagged their books "London" thinking that they would connect exclusively to books about London, Ontario would obviously be a little crazy.

Obviously, some will disagree with this and say we have corrupted the data. The problem is, we can't accommodate both views.

38> In the process, we create a few inappropriate connections between books tagged "London" where the tagger meant "London, Ontario" with books tagged where the user meant "London UK". This is mitigated by the fact that we have done a lot more good than harm and also that anyone who tagged their books "London" thinking that they would connect exclusively to books about London, Ontario would obviously be a little crazy.

You said it much better than I could.

#38: Again, we have a threshold here. One of my issues is that if I tag something "London UK", I obviously want to keep it separate from those London tags, but you're combining it anyway. And in the London case, you're not gaining much; you're destroying the "London UK" tags and not adding much to the "London" tag.

In any case, jjwilson is also arguing that "ray-gun" should be combined with "ray gun", even though the vast majority (100%) of its uses are to mean (Ronald) Reagan.

>37 aulsmith:
My understanding was that it was just those idiosyncratic, sub-cultural references it was supposed to be good at preserving.
So the question - is Reagan = ray-gun as opposed to raygun or ray gun really a folk usage or a happenstance? If anyone cares, I have found instances of both Ronald Raygun and Ronald Ray Gun out on the web.

Though I should also say that I don't think it much matters in this case because the tag has only been used once.

>38 marq:,39
anyone who tagged their books "London" thinking that they would connect exclusively to books about London, Ontario would obviously be a little crazy.
It has nothing to do with the expectations of the taggers, but rather the reality of the social usage of the tags.

London has been used tens of thousand of times, almost always meaning London, UK. Uses of various more specific tags for that city are in the hundreds. Any tagging instances of London with other meanings will be connected to an overwhelming majority of tags meaning London, UK regardless of what combining is done. And the tag list under London+(England) may be pure, but it is a pointless purity in that it excludes the overwhelming majority of tagging instances meaning London, England.

And why should I be forced to use London, UK for my tags instead of London when I and everyone I know would assume that London means London, UK. For those Ontarians who might think of London, Ontario first they must at least be aware that for the majority of the world the assumption would be otherwise.

#41: So the question - is Reagan = ray-gun as opposed to raygun or ray gun really a folk usage or a happenstance? If anyone cares, I have found instances of both Ronald Raygun and Ronald Ray Gun out on the web.

As I said elsewhere, Reagan is more likely to come out one word, be it Ray-gun or Raygun whereas ray-gun as the name of an object is bad English; you don't see that many noun-noun constructions in English as opposed to noun noun or nounnoun.

And the tag list under London+(England) may be pure, but it is a pointless purity in that it excludes the overwhelming majority of tagging instances meaning London, England.

But it includes all the major ones. Giving up a large enough tag list that's pure to enlarge a huge tag list that's not so pure is not the obvious tradeoff you imply it is.

>43 prosfilaes:
Large enough for what exactly? The "London (England)" tag is small enough that only one book has the tag applied more than once.

#42: You're not forced to do anything. Even without London UK merged in, London is still a huge tag list. On the flip side, why should someone else's choice to use London UK for their tags be overridden by your desire to tag tersely and yet use their data?

#44: Fix your filters. If you have "Most often tagged" selected, 1984 by George Orwell (8 times) is at top. You must have "Weighed" selected.

>46 prosfilaes:
Thank you for the correction. But the question remains, useful for what?

"Usefulness" of index points in information theory is generally defined as the signal-to-noise ratio of the material indexed in relationship to the users question. Therefore, it changes with each users question.

However, we know that, in general, even the most determined researcher will generally not look at more than about 100 data points and the casual user will look at much less (say 10 or 20). Since both London UK and London have way, way more than 100 entry points, they're both likely to be pretty useless.

#47: Useful for finding books about London (England). It's not an exhaustive list, but it's still a big one.

On the flip side, what is combining London (England) into London useful for? London is a huge tag already; what's the value in adding a tiny tag list to it?

>49 prosfilaes:
But what makes it useful as a separate tag? The main London tag is far more representative of books about London on LT.

Most combining involves adding a fairly little used tag variant with a more widely used variant. If that is pointless then most tag combining is pointless. The advantages of any tag combining include:
•Making the tag more complete; some of the books on the smaller list are not on the larger
•Improving the ranking and weighting of tag page
•Increasing the related tag linking
•Reducing the clutter on site-wide tag lists by combining tags with the same meaning

50; If that is pointless then most tag combining is pointless.

That's certainly my thinking on the subject. The only utility I see in site-wide listing of tagging is the possibility of small groups of people using similar idiosyncratic vocabularies being able to find each other. To the extent that tag combiners actually combine things with the SAME meaning and let the little gems of odd meanings shine though, it's useful, but once you stick Ray-Gun meaning Reagan in with over a hundred fiction books featuring energy weapons, it becomes pointless.

#50: But what makes it useful as a separate tag? The main London tag is far more representative of books about London on LT.

All Londons.

Most combining involves adding a fairly little used tag variant with a more widely used variant.

Most combining involves adding two tags that are fairly uncontroversially the same. It's entirely clear that adding "role-playing games" to "roleplaying games" improves things, that we should never want to see both "role-playing games" and "roleplaying games" on the same tag list, etc. I certainly would want to see "London" and "London UK" on the same tag list.

Basically, any case where you're combining one tag with another tag the size of London, all you're really doing is cleaning up clutter. In cases where it is just clutter, that's fine, but if people want the original tag separate, if it's more specific and unambiguous then the larger tag, then the value of combining is questionable.

One point that's been overlooked is that if you combine tags you don't lose the individual tags. You can click on the list of tags that's been combined with the main tag and get a tag page for the sub-tag. It claims that there are zero users for sub-tags but it still has tag clouds and books most often tagged with that tag in relation to the sub-tag.

51> You've left out uses having to do with recommendations.

>52 prosfilaes:
All Londons
Of London, UK.

I do appreciate that you are arguing from from usage and not just semantics. But I still honestly don't understand how one would find any site-wide LT feature usage for the London (England) tag that could not be better served with the main London tag. Nor can I imagine how any site-wide LT feature usage would not be better served (if only marginally so) by having all of the tags meaning London, UK together. If I could imagine such a use I would likely cease my argument.

52> ...but if people want the original tag separate, if it's more specific and unambiguous then the larger tag, then the value of combining is questionable.

If people want to use both London and London UK as tags in their library combining them does absolutely nothing to hinder their use of those two tags in their library. Even their personal tag cloud will show those two tags as separate.

#54: By that argument, novel is the most representative tag for novels set in London. It has a lot of stuff that isn't London based, but it has all the London novels.

But I still honestly don't understand how one would find any site-wide LT feature usage for the London (England) tag that could not be better served with the main London tag.

It tells you that the work is definitely connected to the UK London.

Nor can I imagine how any site-wide LT feature usage would not be better served (if only marginally so) by having all of the tags meaning London, UK together.

The London tag doesn't _mean_ London, UK. That's the whole problem. If all the uses were of London, UK, then we probably wouldn't be having this discussion.

I've given up on using the tag pages for anything but checking on combination proposals because so many of my tags have been combined in ways that I find upsetting.

Anyway, obviously I need to change my London UK tag to something that is NOT combined into London.

>54 eromsted:
By that argument, novel is the most representative tag for novels set in London. It has a lot of stuff that isn't London based, but it has all the London novels.

That's just silly. The overwhelming majority of books tagged novel are not about London, UK whereas the overwhelming majority of books tagged London are. Also, the latter is the tag that the overwhelming majority of users pick to represent the concept London, UK. This is clearly not the case for novel.

It tells you that the work is definitely connected to the UK London.
How does this help you do anything? Books tagged London are 99% or more likely to be in reference to London, UK. It is a far more complete list. And any books not about London, UK will almost certainly be readily identifiable from other clues.

The London tag doesn't _mean_ London, UK.
Oh, so it is just semantics. I still don't see any practical use here whatsoever.

>57 MarthaJeanne:
Why do you care? What impact could it have on how you use the site? This is not a rhetorical question. I really do not understand.

#58

At what percentage do you think that the majority becomes overwhelming and combination should be done?

Devil's advocate - SF combining with science fiction? SF is overwhelmingly used to mean science fiction.

* note for people who haven't read my thoughts on this subject before. I am firmly on the side of splitters in this matter. I don't even like that Science Fiction is combined with Sci-Fi (there are significant differences in connotation between the two).

53: 51> You've left out uses having to do with recommendations.

Right. Now the person who used Ray-gun for Reagan will get recommendations of novels with energy weapons. I'm sure that will be very helpful.

>59 andyl:
At 99% usage of one meaning for an ambiguous tag and 100 to 1 ration between the ambiguous tag and more specific variants I would definitely combine. At 90% and 10 to 1, I would likely combine, but I would want to play around with the case a bit to decide. More mixed up than that and I might vote to combine, but it would be far less likely. I couldn't say without looking at a specific case.

SF and Science Fiction are difficult because the conceptual meanings are overlapping, but not identical. Looking at a list of books it's hard to know whether SF means science fiction, sci-fi or speculative fiction. The lists generated by those terms may have interesting distinct patters even if they overlap substantially. If the only issue were San Francisco and other such I would combine because those are almost certainly below the 1% range.

London, UK is not conceptually overlapping with London, Ontario. There's no question of London being an interesting, distinct concept. The only questions are signal strength and noise ratio.

What's the difference in connotation between science fiction and sci-fi? I would appear to have missed it.

I don't know that I buy into it but here's one summary of the debate:

http://www.pajiba.com/trade_news/scifi-vs-science-fiction.php

Gernsback in his 1926 Amazing Stories coined a term he called "scientifiction". I have also seen "scientific fiction" used for stories like Verne which use a lot of science in them. There are also terms like "Hard SF" which is intended to evoke a more realistic and rigorous application of science to the stories (at the least they must be self consistent) than things towards the "fantasy" (magic) side of the scale.

I run into this with time travel stories which can be "hard" with Einstein-Rosen bridges in black holes or spinning Tipler cylinders to magic rings and portals to knocks on the head or no explanation at all.

Probably most catalogers and taggers are not so specific as others in their application of tags.

I see this thread and others like it that split hairs about combining or splitting of tag terms. What is the result of too much or too little combining? Why do we care about it?

James

#62

See the SF Encyclopedia - http://www.sf-encyclopedia.com/entry/sci_fi and http://www.sf-encyclopedia.com/entry/sf

Discussed loads of places - http://www.sffchronicles.co.uk/forum/535318-sf-vs-sci-fi-a-poll.html and a number of discussions here not least http://www.librarything.com/topic/7041#68689

The end result was the splitters, those who do see a difference, lost out. The masses imposed their will upon us (well they outvoted us - the same combination proposal kept reappearing) which will always happen if there is a small(ish) in-group who do see a difference but the masses do not.

#63 Why do we care

http://www.librarything.com/wiki/index.php/Tag_combining

We care because some tags encode differences in perspective and/or identity. Now, you may not recognise (or care) about those differences but for the people who do they are important.

The best I can hope for is that we are self-consistent in tagging within out own catalog. I think it hopeful beyond words that others will have the same nuance I might have for terms.

My questions came back to thoughts like:

If tags are combined and I use two of them, does it change my catalog or my application of the tags?

Is the tag combination manifested in larger aggregations of tags across the site?

I don't disagree that the distinctions and discussions around them are important. However, there's a limit to how much can be resolved, especially by consensus.

If my tags are untouched in the way I use them, it should not be a really big issue. If they alter my data, it goes against one of the higher-order philosophies behind LT. Given that, I have believed that any tag splits or combinations would have little impact on my catalog. Am I incorrect about this?

Always willing to learn new things about one of my favorite sites,
James

66> If tags are combined and I use two of them, does it change my catalog or my application of the tags?

Absolutely not.

63 & 64: Thanks. I guess I may belong to those youngsters who don't see much difference, tho "sci-fi" isn't a form I commonly use: the tag I use on my own books is "science fiction", regardless of literary ambition or hardness.

(Actually, originally I used "SF", but I decided that was too ambiguous.)

64: The end result was the splitters, those who do see a difference, lost out. The masses imposed their will upon us (well they outvoted us - the same combination proposal kept reappearing) which will always happen if there is a small(ish) in-group who do see a difference but the masses do not.

Yes, exactly. But you will never convince splitters to be lumpers and vice versa. It is personality. The problem is a voting system that creates winners and losers. A voting system that eventually will enforce an intellectual monoculture on LT.

At least on tag voting, the super-majority rule results in deadlock. The alternative is that the splitters (probably) win and the lumpers leave (or vice versa). Where does that leave LT?

It is too late for images. The super-majority appears to takes the view that author pages are like wikipedia pages. (I don't know where that leaves pseudonym authors. I suppose we can set the common knowledge Gender to Unknown and everything else blank). Anyone who thinks otherwise gets their contribution voted off by "Helpers" who will gladly ride LT into oblivion rather than give an inch on their opinion.

I'm somewhat puzzled by your reaction. Yes, author pages are more like wikipedia pages than like facebook pages. Any LT user coming along can change anything (s)he likes - we can only hope people do so responsibly.

Also: what you now call "intellectual monoculture" I call data quality. As I see it data quality, though it's not perfect here either, is LT's best distinguishing point. Do you really prefer to spread misinformation?

p.s.: Don't get me wrong, I not denying you your point of view. But such things are precisely why I think sites like Shelfari and Goodreads are useless.

I have never said that authors pages should be like facebook pages. I am saying that their are a diversity of views about how author pages should be used and all views should be accommodated and respected as far as possible.

You say: "Any LT user coming along can change anything (s)he likes - we can only hope people do so responsibly". But, you and a few others constantly flag and vote off pictures that people add to author (and venue) pages. The LT users that add those pictures obviously believe that those pictures are relevant and appropriate and are a valuable contribution to LT. They are acting responsibly. You may disagree but you have to respect their opinion about how they use their site. Otherwise, you promote only one opinion about how LT is useful to its members. Without a diversity of opinion, LT dies.

But, you and a few others constantly flag and vote off pictures that people add to author (and venue) pages

And others who oppose the flagging and voting off are quite welcome to cast their votes as well. That's the point of voting -- the decision isn't made by any individual. No single opinion outweighs any others.

There is a pretty clear majority of users that agree on what author and venue pages should be. You disagree, and that's fine, but you too must respect the opinions of others.

It is the great illusion of voluntary surveys and in this case voluntary voting that the result somehow represents a majority view. It is only the majority of the people who vote, not the majority of LT users.

This statement: "There is a pretty clear majority of users that agree on what author and venue pages should be" is completely false. What is clear is only a majority of the people who vote.

What is also clear to me is that the people who flag images don't even bother to dicuss the issue with users who upload the images. I do. Many of them think that BarkingMatt, Lilithcat etc. are LT staff and that they have done something wrong. They don't understand that they are just other users. They don't even both to vote to keep the picture they have uploaded. They also come to negative conclusions about LT.

(Images voted off LT often appear on Goodreads author pages if they are not there. And that is only partially because I put them there.)

What is also clear is that the solidification of LT into an intellectual monoculture is something that has progressed over time. There are plenty of interesting, relevent, appropriate, beautiful images on LT author pages and venue pages that if uploaded today, would be voted off. This shows that earlier in its life, LT was a far more diverse, creative and open community that it has become today. I always hesitate to give examples of this as someone will probably flag them but:

Author image (Beethoven):


and there are many venue pages that include images of the person the library was named after etc..

And as for Goodreads, it is true that Goodreads does not compete head to head with LT. The sites have a different focus and I am a member of both. Perhaps though people should have a look at Goodreads author pages these days. They are well laid out and usually more comprehensive than LTs. Certainly Google usually links to them before LT. Without text (e.g. in the short bio field) and certainly without images, why would google link to them?

You disagree, and that's fine, but you too must respect the opinions of others.

You can't be serious. I respect the opinions of people who want to use the site in diverse and creative ways. I do not respect the opinions of people who think that only their way is right and destroy the contributions of other users.

If you have a democratic system, and you don't vote, can you really complain?

Well I think so, if you don't know the election is on and only 8 people (including I strongly suspect in one case, her band of sock puppets) decide the result, you might think about complaining.

Anyway, people don't complain, they leave.

I get the point about googling. Fine. But I don't think googling that leads somebody to misinformation - showing, for example, the author's favorite pet, potted plant, whatever, anything but the actual author - would leave a good impression.

As for the Beethoven picture: I've always thought monuments dedicated to authors are okay.

I do think the difference between Goodreads and LT goes deep though. To quote Goodreads: "Are you an author or a publisher? Gain access to a massive audience of more than 13 million book lovers. Goodreads is a great place to promote your books." LT on the other hand explicitly chooses not no be an advertising medium. Core business seems to be LT for libraries, in other words: information.

But we can probably argue about what's relevant info till we're blue in the face. We simply disagree. No problem in itself.

And for what it's worth: I do frequently bother to discuss things with the users who upload these pictures though, and when doing so I do try to make absolutely clear I'm not staff.

But I'm getting tired with this discussion. Fine, I'm giving up on this. Have your potted plant authors.

I've seen several people say that they found LT through googling a book or author.

Re Aitken: LT has a disambiguation page with 8 authors. Goodreads doesn't seem to mention more than one. The LT page for the Zen master has a lot of information, including his date of death. The Goodreads page suggests that 'Robert' should be invited to join Goodreads. Yuck! if you must, then at least use the full name and not just 'Robert'.

Your arguments about images are not useful because the only specific image you give people a chance to judge is on the page.

I think I probably originally found LT through Google too.

The Robert Aitken example is very strange and I think someone on LT should investigate it. Why doesn't Google find any LT page when searching for Robert Aitken? I wonder if the disambiguation somehow derails Google's web crawling algorithms. It does find a link to LT when searching for "Robert Baker Aitken" on page 27 but strangely nothing for "Robert Aitken" in the first 50 pages. Before then, many links to Goodreads etc..

Hopefully that is an anomaly but it is pretty normal that Google finds Goodreads before it finds LT. The main problem I think is that LT's author pages rarely have a block of text e.g. that should be put in the "short biography" field. Images seem to increase the rank too. What does it mean for LT's future if you don't come across it in Google before you find something else?

As for the kinds of images that get flagged, for some reason (perhaps this topic), there has been a sudden cleanup of the currently flagged images. But I think flagged images fall into a few categories:

1. exact duplicates. - a waste of space but who cares? In my opinion, should be deleted.
2. errors. Due to error or author separation, an author image is not related to the author it is listed under. IMHO - should be moved to the correct author, flagged for removal with a note.
3. errors. An image added to a venue that has been added to the wrong venue. Never seen it and probably wouldn't realise if it happened. Should be moved to the correct venue and flagged with a note.

All fair enough. But then:

4. Images relevant to the author they are listed under but not a picture OF the author. e.g. monuments, places where they lived or that are significant to their writing etc.. IMHO - valuable additions to LT.
5. Images directly related to a venue but not OF the venue. Forthcoming and previous events at the venue, pictures of the person the venue is named after or commemorates etc.. With the new events interface, I found that event related images can now be added to the event listing with a HTML link. So I have been rescuing the flagged images and putting them in there after discussing it with the user. IMHO though, the images however remain relevant to the venue and should stay there. (Note, the HTML link solution is not ideal. There is no restriction on the width of the image so you could easily link a 1000 pixel wide image, destroying the layout of all the LT pages (I think 4 types) that display events. I recommend adding width=300 to the img tags if you do this.).
6. Published images for pseudonymous authors. Authors writing under a pseudonym may use fictional images to represent the persona of the author. These images may not be a picture of a real person and may not be a picture of a person at all. The question is, is the author page of a pseudonymous author about the pseudonymous persona or about the real author? What do you put as the gender of Elizabeta Brooke writing erotic fiction for women? What evidence do you have that the real author is female? IMHO - the fictional images are appropriate and should stay there. In most cases, they are the only images that we are going to get.
7. Pictures of the author's favourite pot plant. Never seen it but always causes concern. If the author's published works make significant reference to a particular pot plant, the image is a valuable addition to the author page (perhaps with an appropriate comment). Authors using the author page to upload images of purely personal relevance? Don't know - never seen it.

Staff has stated clearly and specifically that venue images should be OF the venue, and that author images should be OF the author.

See
http://www.librarything.com/topic/145647#3741735
http://www.librarything.com/topic/145647#3742706
http://www.librarything.com/topic/145647#3748192

> 78: Potted plants: hyperbole. But take a look at http://www.librarything.com/gallery/flagged/removed. There are pictures of cats and toads there. I'm pretty sure none of them ever wrote any books. Where do you draw the line? Maybe bratwurst with sauerkraut is the author's favorite dish. Does that make a plate of such a valid author picture? I think not.

But as I said, I'm tired of this. If my efforts to improve / maintain data quality offend that much I'm withdrawing from the field. As far as I'm concerned you're getting your way. Enjoy, but it's your way. Now that's what I call monoculture.

(Edited so as not to point to the wrong message.)

>78 marq:
Why are you engaging in this discussion on this thread when you already had a perfectly good thread that was directly about this issue?

> 80

If my efforts to improve / maintain data quality offend that much I'm withdrawing from the field.

Please don't.

If I see a cat or a potted plant as an author picture I will flag it.

How did a discussion about making separation of tags devolve into what is and is not appropriate to post regarding pictures?

It's a law of nature it seems: discussions on LT tend to digress.

I'm late with this because I assumed it "must" have been on another thread and so, had a hard time looking for it.

> 71: I have never said that authors pages should be like facebook pages.

Sorry but your comment in #17 made me think you did: "because the image they use to represent themselves on Goodreads, Amazon, Facebook and/or their blog is not good enough for LT".

Amazon, Facebook and blogs are advertising. Apparently (see quote in # 76) so is Goodreads. LT isn't.

84> not really a digression. The thread is about absolute voting structures on LT and how they promote an intellectual monoculture on LT. The topic referenced above about the threats because LT is 40% owned by the company that owns 100% of its main competitor (Shelfari) criticises Shelfari for being a monoculture and aims for LT to be more open and diverse. However, as we see, LT is anti-diversity of opinion and its voting structures enforce that. I think that is a threat to LT's survival.

I did however digress on the examples of images that would normally be removed from LT in response to MarthaJeanne's criticism (77) that the example I gave above was not comprehensive.

79>Staff has stated clearly and specifically that venue images should be OF the venue, and that author images should be OF the author.

Yes one of them has. You think they have the final say on this? I assume you will "help" by flagging any image you see on author or venue pages that are not OF the author or not OF the venue and put them to the vote. You can start with the example I gave above for Beethoven.

#87

UM NO.

This was about adjusting the algorithm of the separating of tags. NOTHING to do with the overall voting concepts. Please don't bring something that has been spoken of on multiple threads to one that has nothing to do with it.

To clarify: "Recommend Site Improvements: Make separating a simple majority".

I oppose. In my opinion, decisions on LT (all decisions) should be a consensus of a diversity of opinions. A majority, even a super majority is not a consensus. Without a consensus of a diversity of opinions, LT becomes an intellectual monoculture. Precisely what Tim Spalding has said LT should not be to differentiate itself from its competitors.

I'm assuming you wanted a clarification and that you were not just trying to shut up someone you disagree with.

#89: Without a consensus of a diversity of opinions, LT becomes an intellectual monoculture.

Consensus frequently forces things to be a monoculture. If there exists a person with an absolute veto on anything they disagree, there doesn't tend to be much dissent; letting everyone be that person at the same time doesn't help much.

A consensus is a decision that everyone can live with. If an individual vetoes a decision, it is not a consensus. In tag combination, the practical result is the way books are sorted in the list of books tagged. As tag combination never actually changes users tagging, LT must first count the number of books with individual uncombined tags and then aggregate the totals for the combined tag group. The books are then sorted in descending order of the aggregated total. Instead, the aggregated total could take into account the strength of the association based on the number of for and against votes. So for example, if 50% are for and 50% against, only 50% of the books tagged count towards the aggregate number used as the sort order. I am simplifying and I realise we are talking about a cartesian product but I have done something similar in my work (relating to insurance fraud detection) with a lot more data than LT has.

So tags are not combined or separated, they are associated strongly or weakly based on user's votes. Similarly, images are not removed, they are promoted, demoted or even hidden (requiring an extra click to view) based on people's votes.

Of course, better than a system solution is a change in culture.

#91: A consensus is a decision that everyone can live with. If an individual vetoes a decision, it is not a consensus.

I don't see how that has anything to do with reality. It would be wonderful if everyone could agree, but it rarely happens, and when it doesn't, consensus rules mean that someone, anyone, can stop.

Instead, the aggregated total could take into account the strength of the association based on the number of for and against votes.

There's no consensus for that. Sorry.