Anyone want to have a relationship?

So, Chris is planning relationships between works. The sort of relatonships we have in mind are that Grendel is a modern retelling of Beowulf, that Barry Trotter and the unauthorized parody is a parody of Harry Potter, and that Walbank's A historical commentary on Polybius is a historical commentary on Polybius' Histories (The rise of the Roman Empire being the Penguinification).

The basic question is this:

Do we define some relationships or leave it free-form?
Are there special types of relationships—ones that by necessity cause a corresponding relationship from the other direction, for example?
What is the potential? What are the pitfalls?

Is this global and shared, like CK, or local and aggregated?

Assuming it's common, I'd say don't pre-define the relationship types, because anyone can go back and standardize them as a consensus arises in the Relaters group.

This sounds a great idea! I think it would be good to have some definitions, but whatever the definitions are, people are going to use this feature the way they like - many people don't read guidelines or definitions.

One relationship that came into my mind is biographical works and fiction about the same person. Relationships between fact and fiction.

Will we eventually be able to have contains/is-part-of relationships? What about between the original and it's translations?

>5 jjwilson61: A translation and its original are almost always the same work.

Under current definitions yes, although there is that troublesome dead language exemption. But then "same work" is just another way of saying that two editions are connected with the "same" relationship. Allowing us to connect them with the "translation" relationship would allow a more subtle description of the real semantic web of literary objects.

I forgot the real reason to have a translation relationship. If you separated all the translations out but added them back in with a translation relationship then Tim could program in the dead languages exception (that is he would change the code to treat translation relationships as if the works had been combined except in the case of dead languages).

This would eliminate one frustration of combiners which is that many people don't understand the dead language exception and make combinations when it isn't appropriate, which combiners have to separate, if they even notice it. My understanding is that the Harry Potter Latin translations keep getting combined with their modern language counterparts and I don't think its the only one.

Once the translation relationship is in place then Tim could add the contains/part-of relationship which would eliminate a whole other, larger, class of inappropriate combinations.

Let's leave translations out of it for now, okay?

Is this global and shared, like CK, or local and aggregated?

It would be a field in CK—or a set of fields.

Some ideas:

Work->Work:
Commentary
Parody
Discussion
Retelling
Argument against

Work->Author
Biography of

I would rather put this kind of info in tags.

Getting the relationships straight inside works seems much more important to me. There are a lot of cases of revisions, and different translations, and bi-lingual editions (with or without the dead language case) that cry to be able to both have them together and separate at the same time. Whether or not it's what you want, you are going to have these books entered into the relationships area, because we have no other way of doing it right now.

If you are going to do relationships of the variety you mention, are there any that shouldn't be made to be visible in both directions? If A is defined as a parody of B, that is important to see about A, but if I don't already know about it, I'm more likely to see it if I'm looking at my book's page and someething shows up to say, 'A is a parody of this book.'

"contains/part-of" seems like a really important one -- unless it's going to be handled some other way. It does seem different from the other examples that have been suggested. (And actually it looks like it should be a catalog feature, rather than a CK feature, so yeah...)

"cites/is-cited-by" (fiction as well as non-fiction)

Ideally I'd like to see a distinction between, for lack of better words, anthology-containment (the relationship between Hamlet and The Complete Works of Shakespeare) and volume-containment (the relationship between The Fellowship of the Ring and Lord of the Rings).

Definetly dual way, auto by default. In rare instances when it may no be appropriate maybe allow editing.

Really really want a "is contained in" / "this contains" relationship.

I am not sure this idea of building relationships between works makes a lot of sense. It would indeed significantly overlap with other LT functions:
tortoise's example looks to me as if it would be covered by the series functionality.
Most other cases could be covered with the tag functionality.

I however like the idea to link a work to an author to get its biography, although this could be managed by the CK "characters" field too...

Lumping containment into the series functionality makes it difficult to use to draw Zeitgeist-y conclusions. For example, the number of people who own The Fellowship of the Ring is very close to the number who own Harry Potter and the Goblet of Fire--it's just that half those people own it as part of The Lord of the Rings. It'd be nice if you could see that directly, and no analysis of series data will let you do that.

Similarly, while you can tag things as parodies, commentaries, or whatever, it doesn't let you go to a work page and say "LibraryThing, tell me all the parodies of this book." Which seems like kind of a neat feature to me.

I don't have any interesting suggestions, but I do like this idea.

Only if you'll respect me in the morning.

>18 lilithcat: Zing!

But just how closely related will the books have to be?

This is similar to the idea I mentioned in the 10 things thread, showing criticism or commentary on a book. I think it's a great idea.

But I think we need to keep separate those relationships that should be handled in CK and those that shouldn't. Part/whole relationships shouldn't, because those should affect sharing; same for translations (which are out of the discussion anyway). Cites/cited by, discusses, responds to, criticizes, things like that are more appropriate for CK.

Also, the suggestion of "biography about this person/fiction about the same person" should already be handled in CK by the character field.

>11 MarthaJeanne:, 16

You cannot tag other people's books...

If "relationships" includes "contained in"/"contained by" information, I am 116% all for it.

Otherwise, I'm not opposed, per se, but I do think that it's got the potential to duplicate the information in CK, member recommendations, and elsewhere.... And while you can't tag books you don't have in your catalog, searching for the "harry potter" or "pride and prejudice" as a tag pulls up long lists of related books already. The exception would be obscure books and edge cases, in which case... why not make it a member recommendation?

I think the point is to be able to not just say that the works are related but how they are related. And I don't see how tags can be used to connect two and only two books.

> 22 duplicate the information in CK

Back in #10, Tim clarifies that this is a part of CK. So, it's pairs of fields, one the relationship type and the other a work / author link.

Now, another approach would be to (1) allow touchstones in CK and (2) make touchstones more reliable everywhere. But Tim is on record as not liking the first part.

And a little bit more structure allows the automatic inverses suggested in #11. With touchstones, all you could have is “what links here.”

Great idea, but with the caution that nperrin brings up in #20:
But I think we need to keep separate those relationships that should be handled in CK and those that shouldn't. Part/whole relationships shouldn't, because those should affect sharing; same for translations (which are out of the discussion anyway). Cites/cited by, discusses, responds to, criticizes, things like that are more appropriate for CK.

"Part/whole" isn't really a work-to-work relationship anyway (it's an, um, part-to-whole relationship), so it shouldn't be included. For the same reason, I would not include the work->Author (biography of) that Tim suggests in #10. (Seems to me this last one could easily be found using a tag or title search anyway.)

I would definitely recommend trying this with limited options at first -- those Tim mentions in his list under Work->Work in #10 are a good start.

I would definitely not allow users to enter other (free-form) kinds of relationships, at least not at first, because I think it very likely that this would result in a misuse of work-to-work relationships that would (a) reduce the quality of data and usefulness of the feature, and (b) result in duplication of other LT functions (as some here have already pointed out). After the feature gets going, Tim can always add other kinds of relationships based on feedback from the LT community; but once you let the free-form cat out of the bag, well...

P.S. Edited to add: personally, I would add the translation relationship. There are some strong reasons to do this. But I understand that translations are off the table for now.

I would make relationships one way and have a predefined list.
in general the relationship would be from the dependant work to the main work. i.e. commentry on, parody of, retelling of.
Then have a page for each work, related books, where the inverses have been calculated giving a lists of commentaries, parodies, etc.

The problems I forsee:
1 Biblical commentaries
a) do they link to each version of the bible (possibly depends on if the commentry is stongly linked to a particular version)
b) commentaries on each book (e.g. The Gosplel of John) would get linked to the bible as a whole.
2 Retellings of folk/fairy tales and mythology - how is the "Original " defined. I am sure that several of us would like to be able to find all retellings of say Beauty and the Beast. (a title search gives 285 works!)
3 Arguments about the boundry between parody and retelling e.g. there and back again
4 A contains relationship might cause a large influx of short fiction so that links coulds be made.

I don't see the problem with a contains relationship working this way, but I would like to see such a relationship affecting recommendations. If I have the complete works of Shakespeare don't recommend Hamlet prince of Denmark (which is included). Going the other way is more difficult. If I have say The Lion the Witch and the Wardrobe but no other Narnia books then it is reasonable to sugest the Chronicles in a single volume, which would be cheaper to buy than the other six books. But if I have six books, recommending the seventh would be better than the one volume edition.

I'd like to see continuations that should not be covered under CK series as a relationship too. I'm thinking particularly of the "sequels" to Pride and Prejudice or something like the recent ER book Becky to link to Tom Sawyer.

Or Gone with the Wind and The Wind Done Gone.

Drat. I thought LT was opening a dating service. :P

I like the idea of this tool, users will be able to follow a spider web of related books. The subject binger in me will love that!

Count me as another voice who loves the idea. I think having some pre-defined structure will make the feature more useful than just free-form, and I think the part-whole issue really is of a different sort entirely, but that's the one I'd like to see the most.

It would be kind of cool to mimic the idea behind literature map. To demonstrate for those that haven't seen it, you enter an author's name and a jumble of names pop up - the closer the name, the more likely you'll like both authors:

http://www.literature-map.com/alan+moore.html

I bring this up because there are really interesting relationships, and then there are relationships that may get murkier. In my head, I'd kind of like to see a similar map or a web where the closer the relationship, the more likely I am to be looking at retelling/concordance/commentary/capitalizing on a franchise relationships and the further away the relationships, the more we're looking at something the same but getting out there.

In my head, I might be looking at a center with The Wizard of Oz (and understanding that the series are not relationships, but understood, right?), move to The Wonderful Wizard of Oz: A Commemorative Pop-up as a close 3-D retelling,

move out a bit to Wicked as a retelling,

move a bit further out for The Wizard of Oz and Philosophy alongside The Historian's Wizard of Oz: Reading L. Frank Baum's Classic as a Political and Monetary Allegory for additional commentary,

further out to The Lost Girls as a retelling that includes Dorothy,

moving further out to Who Stole the Wizard of Oz as a book where our original book is used in the plot

and finally out to The Wizard of Oz and Other Narcissists as a book that merely works on themes...

I see this in my head - it almost works as a spiral out - and like the literature map, clicking on one relationship would move the spiral out to a whole new set of relationships...

27, 28.
Relationship "Unauthorized sequel"?

You really have two extreme choices with all kinds of compromises in-between.

Choice 1 is to follow what has been done by "Library Scientists" for almost a hundred years now and use a fixed ontology of relationships. Library Science has fallen out of favor but a fairly elaborate collection of relationships between books and other published material has been defined. The Library of Congress, for example, uses some of them in its catalog entries. These are not keywords but relationships -- most of them binary, but there are n-ary relationships as well.

Choice 2 is to do what some computer systems have done (The Lightweight Directory Access Protocol or LDAP, for example) and that is to support a mechanism for creating new binary relationships. A whole bunch of these are predefined but you can specify a number of your own as extensions.

In-between, you have options like XML that allows you to define a schema with fixed relationships but allows you to include other undefined attributes in the data. These won't show up but can get saved until the system decides on how they should work.

IMHO, doing a subset of what has been defined by library scientists for the Library of Congress would be reasonable.

These include Anthology, Sequel, Collection, Commentary, Translation, Re-telling, and a few others. However, some of the really creative ideas of #31 would not be there!

hmmm...I guess Tim's become one of those guys that's just not into the "whole relationship thing!"

=P

This may come under "retellings", but some way to denote the relationship between a classic and its simplified version meant for children. Or, I guess, any abridgment.

And certainly some form of "based on the series/character", "inspired by", or "unauthorized sequel" to cover modern authors continuing classics - like the Pride and Prejudice sequels mentioned above, or recent Sherlock Holmes novels.

I'd like abridgment and retelling, as the two can be very different.

I'm like something about anthologies and omnibuses
recognising that one story is the same as another, even though the second one is published with a third different story. See Nora Roberts

Did this make it into some plan or not?

I think the potential is enormous if your goal is to find works that might interest you because you liked something else. The main pitfall is that it gets out of hand, becomes a mess of ill-defined relationships, and the noise overwhelms the data. In order to avoid this, a fixed taxonomy of relationships seems a must.

On the subject of cardinality, I think that making them bi-directional is the best bet. More interesting to me, however, would be the transitive nature of relationships. For example, knowing that McKinley's Beauty was a retelling of something de Beaumont wrote in French would probably stop me cold. But, perhaps knowing that Lackey also retold it would prompt me to search for the latter. For that reason, it would be nice to have a page that showed multiple jumps in the relationship tree. The idea #31 illustrated of the literature map was very exciting, though I'd want to have the relationship type discoverable.

In general, I find myself to the loosey-goosey left of most LibraryThing users. I don't think the sky will fall if we provide some standard relationships, but allow users to add new ones.

Tim,

Are you talking about those added relationships being private to the person who added them or shared amongst the wider LT community?

My feelings are that the latter is the better way to go as it may suggest a relationship (either a listed one or a new one) s that wouldn't necessarily have been thought about. The only problem I guess could be the total length of the list of possible relationships. It may become too unwieldy. Maybe if it does a link to a wiki page where we can communally list out non-standard relationships so that we can share them (and also have our imaginations poked with a stick).

>42 andyl:

Public, through Common Knowledge.

I think the list will be set, with perhaps the option to pick up any terms used more than X times, then an "other" option that turns the menu into a text box.

I'm happy to finally see a serious discussion of this feature. Let me definitely throw my voice in with the "contains"/"contained in" suggestion.

I was really gung-ho on this, and wanted to implement it right away. The reaction here convince me there are lots of problems with this, and some caution is needed.

Didn't mean to get all wacky with the spiral idea-thingy...

=/

45: You should do it anyhow with a few fixed roles, then people can gripe or not gripe as they see how it actually works. :)

Oh, there will be griping... :)

Who, us????

Never. ;)

Yeah, like they wouldnt gripe, they would gripe if they were beheaded with a clean sword.

> 45

Tim, the "contains"/"contained in" relationship is different from other kinds, and may be implemented separately. It abstracts the notion of a work one step further from that of a book, which IMHO fits perfectly with the LT philosophy.

Reading the Combine/Separate Works philosophy statement, I see that the goal is to create social links between people listing the same work, regardless of edition. I think this benefit should be extended equally to those who have Frankenstein as stand alone book, as well as to those who have it as part of the Frankenstein, Dracula, Dr. Jekyll and Mr. Hyde anthology.

The difference between the "containment" relationship and the other kinds of work-to-work relationships suggested is this: It could improve the flawed relationship between works and books.

LibraryThing treats the relationship between works and books as one-to-many: a work comprises many books, and a book belongs to only one work. In reality, the relationship is many-to-many: a book contains one or more works, and the same work may be included in several books. At first, I thought that changing this would require major changes to LibraryThing's data model and the unraveling of many book combinations, but this talk suggests something else: If we say that a book contains one work that is everythinginthebook, then the current one-to-many work-to-book relationship could be retained, and the rest could be accomplished by a relationship between works.

This offers answers to a number of problems:
That in message 52 above. There may need to be some simple math to keep that anthology from counting three times.

Substantially different publications that are controversially combined as works (various critical editions of Shakespeare, for example) could potentially be separated while retaining a relationship: my book The Arden Shakespeare: King Richard II contains the work The Arden Shakespeare: King Richard II by Charles R. Forker, which in turn contains the work The Tragedy of King Richard the Second by William Shakespeare.

Short stories and other works that aren't in anyone's library as discrete books could be entered as works contained in the works that are.

The other kinds of work-to-work relationships are of limited use without this one. It seems silly to have a relationship saying that Grendel is a modern retelling of Beowulf if a copy of Beowulf and Other Old English Poems can't participate.

I'm worried that if it's completely open, it will be so ill-defined to be useless.

The contained-in/contains is useful and fairly well-defined. Many of the others, even the interesting ones, not so much.

You say that Barry Trotter and the unauthorized parody is a parody of Harry Potter, which is fine until you look at the touchstones and realize that Harry Potter points to Harry Potter and the Half-Blood Prince, which was published three years after the first Barry Trotter. But it's not simply true that the first Barry Trotter book is a parody of Harry Potter and the Philosopher's Stone, either; if you've had the misfortune to read the book, it's sort of a parody of the entire series up to that point. Likewise with Sherlock Holmes; there's a great corpus of post-Doyle writings, but few of them connect to any specific work of Doyle's. There's a lot of really vague connections here, which seems problematic for the formerly fairly concrete CK.

I agree that it'll work much better if there are defined categories to choose from. With standard categories, the same idea doesn't get called several different things, and you always can add an "Other" at the end of the list.

And this has been said, but I'll put my own spin on it: This feature would be pointless without some sort of contains/contained in category. Many (if not most) people would know that The Fellowship of the Ring, The Two Towers, and The Return of the King are sometimes published in a single volume as The Lord of the Rings. Many users might have observed that Democracy in America is available in a really big volume or in two smaller ones. But how many people would know that Yvain, or the Knight with the Lion can be found on its own or as a part of Arthurian Romances?

Mm hmm. And I would say that two libraries, one having The Ultimate Hitchhiker's Guide and one having each of the books comprised therein, have more in common than two libraries having substantially different editions of Shakespeare or the same work in different languages. Yet the latter are combined while the omnibus has no socially functional relationship to the separate volumes.

55>

I would. ;)

That said, I want contained if only because it will stop me from getting recommendations for omnibi when I own the constituent parts in separate volumes.

Of course, there will be issues.

There will be Annoying People who make placeholder "all the books by Author X" works, and then tell us that everything is Contained In that work. This seems less than useful, but I don't think there's any way around it.

> 57

That's a very good point, lorax. Cleaning up the recommendations of duplicates would be another benefit of contains/contained relationships.

BTW, I don't think that combining works into "all the books by Author X" is useless or even less than useful. If someone creates these combinations, then they probably find it useful. If someone else doesn't find it useful, they need not use it.

>56 pterodactling:
That's part of why I want this category, too. But just having it without it affecting recommendations might be helpful for knowing what else we have contained in what we've cataloged.

58>

I think I'm not understanding.

When you say "I don't think that combining works into "all the books by Author X" is useless or even less than useful.", you're referring to someone making a personal placeholder, right? Not actually combining the works? Because actually combining them goes well beyond useless and into destructive.

> 60

Lorax, sorry, I didn't mean combine in the technical sense. What I had in mind is someone creating a new work, representing all the works of a single author. The individual works would be related to it through the "includes" relationship. Completely opposite of destructive and only possible with inclusion relationships.

What's the point? You can just go to the author's page to see all of his or her works.

63: Yeah, I agree, unless it's something like The Complete Works of Shakespeare.

63> Nathanael West wrote four books. Four. That same Nathanael West has 12 volumes on his author's page (+ some from another author of the same name.) That's a small, if definitive case. The Agatha Christie page has 799 works listed, which is pretty good for an author that wrote 80 novels. (Looking her Wikipedia page, there's maybe another 80 volumes of plays and collections of short stories that would make up her complete works.) The rest is short stories listed independently, omnibuses, comics, various uncombined foreign editions, etc.

Furthermore, the author's page is not set up to show you which books you have and which you don't, unlike series pages.

65: And green checkmarks would be nice there too.

>63 jjwilson61: Some works show up under the translator or editor depending on how the majority of people enter the works. Also, if someone is a co-author of a work, the work might not show up on the author page.

Well, there isn't a Contained/Contains relationship at this time so as long as we're postulating future developments to making the author pages better too. In fact the Contained/Contains relationship should make it possible to distinguish between omnibuses and the works they contain and use that info to clean up the author page. (Should omnibuses be works at all? They are just the combination of previously existing works with nothing new).

I really like this idea. Could we also have a "Reference/Guide" category? Does this mean that we will have a new works page for the Bible and all translations under it? Right now their are just a lot of unrelated works pages for different translations and publications. If all of the dictionaries, encyclopedias, atlases, etc. are to be related, I don't think anyone wants to go to each individual translation's page and enter the same books/relationships fifteen to twenty times. The main work page would of course be huge, but it could save some time.

I very much want contained-in relationships, but that would be a whole different kettle of fish -- because what I want to do with that is include the contents of my anthologies and collections, which ought to have their own separate works-page. So frankly I think they could stand to be part of a reconceptualizing of the present implementation of the title / work concepts, and would include:
* title
* editions of
* personal copy
* translations / adaptations of (perhaps an adaptation into comic form, as with Classics Comics, is best considered a type of translation into visual language)
* included-in, part of series, etc.

But I think relationships as originally proposed could be quite darling.

I also favor a discrete list of relationships at first, although one would have to be able to add multiple such relationships -- for example, X could simultaneously be an adaptation of Y and a retelling of Z. I am ambivalent about user-added relationships -- I foresee users adding in many relationships that are otherwise covered with other LT fields. Maybe, a user-added field could show up, but be greyed out until 5 separate users green-flag it.

Anyway, some important relationships would be:
* Parody of...
* Retelling of ...
* Abridgment of ...
* Novelization / fictionalization / dramatization of ...
* Commentary on ...
* Refutation of ... * Guide to ...

Obviously we needn't reinvent the wheel here.

Reviving the thread so its easier to find by those who have started discussing something similar.

>10 timspalding:.

Tim, what's the difference among a commentary and a discussion and argument against? Couldn't they all just be "critique"? I think that if you're going to start, you should start relatively smal, or do something like you did with the Open Shelves Classification (where you crowdsourced some of it). That way, you can gauge the needs of the user and build it to suit. You could even start with a general relationship tag and allow the users to insert (type of relationship) in the field, and then evolve it from there.

Other than that, I am 100% in favor of work->work and work->author relationships. 110% even.

Would it also be possible to have a character is author relationship for character pages? For example, if Salinger appears as a character in a book, he is listed as a person/character in the CK area of that book. Then, on that person/character page, it could say "Person/character is {author}", and then on the author page, it can say "Appears in the following works: ..."

Please?

Also, I'd love for there to be a "this work contains these works" type of relationship, as others have mentioned.

whatever happened to this idea? it's wonderful.

Yes, especially the contains/contained-in part. Owners of anthologies, collections, omnibuses, etc. would all benefit greatly.

Also, it'd be nice to have something analogous for authors. How many times have you run into a book with author "Last2, First1 Last1 and First2"? That's a multiple author text that's been mistakenly entered as a single author from Amazon or some library's catalog data. There should be a way to solve this problem, just as author combinations have solved the misspelling problem.

So, another bump to this thread!

yep, really would be great to see that contained in thing happen. i know people have been talking about it for a while. i just manually entered an article out of a scholarly publication, and it looks a mess. i need to keep track of it for my own publication research, and that's really the best solution i could come up with. but it's not a pretty thing.

This has been asked for again (here).

Bump? I don't want this to get forgotten about among all the other stuff Tim's been implementing. (Unless he already has implemented this and I missed it.)

Following up on >77 AnnaClaire::

I come to many suggestion / request threads very late, like this one. I end up wondering if there's an easy way to identify (a) the status of the request, though of course not necessarily all items mentioned in the thread, and (b) related threads (such as multiple requests for the same, or merely requests that in some way touch on the original request).

I've read about a development "road map" but don't think I've seen it. Is this posted anywhere? And does it make reference to recommendations / bug fixes as they're addressed or scheduled? Even if the road map were purely historical: I sense it won't make things easy for our LT staff if everyone could critique the projected work schedule ...

Part of my request could be handled with searches, I realise. I've had problems getting searches to work for me, and I understand part of that may be LT functionality since this is an area in development / improvement. But I'm sure part of it is me, too.

I post this primarily as I suspect there's something out there that does part of what I'm asking, and I've simply not stumbled upon it. So much of my time on LT is intermittent (cough, at work) that I don't get a systematic review of how LT works.

You're certainly right about talk search leaving much to be desired (searching for works, while far from perfect, is a bit better).

#78 elenchus There's this

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

but it hasn't been updated recently.

AH! Now that calm reminded me, I do recall seeing that before. Useful.

I also came across the plan for Local:
http://www.librarything.com/wiki/index.php/LibraryThing_Local_Plan

That's pretty nifty. But again, I don't expect it would be useful to do that for LT development, generally.

The more I think of it, though, it would be nice to have a historical list of what was done, and when, and link various threads to it. I suppose I should create a Wiki page for that, if I want it badly enough.

#81 > See
http://www.librarything.com/wiki/index.php/Index_of_features (Index of features) and http://www.librarything.com/wiki/index.php/History_of_LibraryThing (History of LT).

<sweet, little-girl voice>

Could we de-jack the thread please?

</sweet, little-girl voice>

Pardon! I'm releasing all prisoners, apologies all 'round.