Wrong automatic combination

when I add this work Tokyo revengers (Vol. 25) 978-8834919637 to my collection LT combines it automatically with Tokyo revengers (Vol. 1) https://www.librarything.com/work/26376711
I had the same problem with different ISBNs Tokyo revengers (Vol. 16), Tokyo revengers (Vol. 20), Tokyo revengers (Vol. 21), Tokyo revengers (Vol. 22), Tokyo revengers (Vol. 23)...

Not a bug. The automatic combiner makes mistakes. There is no way to make it perfect. I have separated it.

Auto-combining tends to ignore anything in parentheses, or after a colon. (I agree that's unfortunate).

But it really helps not to have to manually combine a lot of books when publishers use different subtitles for hardcover, paperback, editions in different countries... Not to mention the garbage that Amazon often adds to titles. You win some, you lose some.

Agreed. It's a trade off, and it will need regular manual intervention either way.

>2 MarthaJeanne: yeah it's a feature
I had to separate 24 works and combine them with the right work
nice feature

>6 supersidvicious: But you did not need to chase another few thousand editions which are automatically combined despite the slight differences in their titles.

It misfires occasionally but it also works as intended most of the time. It is a balance.

>6 supersidvicious: nice feature

No one else called it a feature (I checked the thread to be sure :-). It is a case where a choice has to be made: If you don't have an autocombiner, then a lot of books, especially the good old ones without ISBNs, will not combine with anything -- and you will lose access to Common Knowledge, reviews, discussions, popularity reports, all those add-on features.

But if you do have an auto-combiner, it will sometimes give false positives. It might be good if the autocombiner had some sort of warning or option in some cases: "There is a colon in this book name; should I combine it with X?" But that has a burden, too: a lot of people will not know how to manage it.

LibraryThing has a solution: the Combiners group. Many people there seem to be happy to clean up problems like this, so users don't necessarily have to do it themselves.

There is no perfect answer. There can't be, given how many books there are that were published before there was any real system of standardization.

>8 waltzmn: Ummm. What about the self published books and Amazon published books that don't bother with those old fashioned ISBNs?

>9 MarthaJeanne: Ummm. What about the self published books and Amazon published books that don't bother with those old fashioned ISBNs?

Not sure I understand the question. Look at my library and you'll see that I have plenty of books with no ISBNs, so I know the lack makes it a lot harder to enter them! And I know the autocombiner makes mistakes.

The flip side is that a relatively small fraction of my books are popular enough to have a high number of editions.

So I'm talking about books with no ISBNs to sort out their relationships, but which also have titles that (to the autocombiner) are ambiguous. For me, having it ask me a question about books with colons would probably save me time, since it would save me a lot of wrong combinations. But if my choices are no autocombiner or an autocombiner which makes mistakes at the current rate, I'll take the latter. I have just had to learn how to combine and separate works. :-)

I wonder if an advanced setting that could be set for each user that told the auto combiner whether to stop at a colon or look at the whole title would be a way forward. That way if you were entering a lot of periodicals you could turn off colon detection

>8 waltzmn: so the auto combiner takes 29 unique Italian ISBNs and unique titles from different works added for the first time in LT and combines it with Tokyo revengers (Vol. 1)
if it's not a bug and it's not a feature how can we call it?
a beature? a fug?

>7 AnnieMod: there's no point in speaking about cases when the auto combiner works

>12 supersidvicious:

A bug is a fixable coding error. It is not simply something a program does that you don't like. If you can produce an algorithm that does better than the current system -- and by better I mean something that is consistently more reliable, not something that fixes your particular problem, no matter how bad it is -- I'm sure LibraryThing would like to see it.

I don't doubt that this is irritating. I can list many times the algorithm has done unfortunate things to my books, too. I have not proposed a better algorithm, because I can't suggest something that will fix my problems without breaking something else.

>13 supersidvicious: You cannot have one without the other though.

Either the auto-combiner only works on exact matches and you end up with thousands less combined daily that we have now or you go with the probabilities and best guesses and you have a certain very low percentage of misfires. It is annoying when it catches one of your books and there are few known cases where it is really bad (magazines and comics and multi-volume works) but as these are not clearly marked as such in sources, there is no way for the algorithm to be calibrated for them specifically.

Is there no way or reason one could not have a "do not auto-combine" option when entering a work manually?

>16 DuncanHill: Is there no way or reason one could not have a "do not auto-combine" option when entering a work manually?

For me, and presumably for you, that would be a good idea. But the other side of the question is, what fraction of users would understand it -- and what fraction of users who don't understand it would create vast numbers of works that should be combined but aren't? I suspect the number of the latter is much higher than the number who would be helped by it.

I'd personally like having the option -- were it buried away somewhere so that only those who know how to use it could choose it. :-)

I think this an important point in any debate about the auto-combiner: What fraction of users understand it? Those who know about this forum are probably not a representative sample. :-)

If entering a book manually, it's already possible to deliberately use a title that lessens the risk of autocombine where it doesn't belong. For example don't include the parentheses around the volume number, or omit a colon if it's not a mere subtitle but an integral part of the work's title. But it gets arguable whether it's more efficient to do this manipulating, compared to just letting the autocombine happen and going back to separate it after. Depends on your workflow/if adding one book or a batch/etc.

>16 DuncanHill:
An override "do not autocombine" tick box is an interesting idea. But possibly having that available upfront for new/inexperienced users would create as many problems on the opposite side of the seesaw: "my book hasn't combined??" "why am I the only person with this book?" "why don't I have any recommendations?" etc. if they've enabled that setting and left it on inadvertently. But could be something useful to have.

>18 Nevov: "If entering a book manually, it's already possible to deliberately use a title that lessens the risk of autocombine where it doesn't belong". - True, I try to do this by omitting colons wherever possible, but it is upsetting to have to use a title which is not that on the cover of the actual book on the desk beside me. But this is an option that is only really obvious to the more experienced user.

"But possibly having that available upfront for new/inexperienced users would create as many problems on the opposite side of the seesaw: "my book hasn't combined??" "why am I the only person with this book?" "why don't I have any recommendations?" etc. if they've enabled that setting and left it on inadvertently". - I would suggest it on a case-by-case basis for manually added books only - and I see no reason it couldn't be restricted to eg members with more works than than 1 SD below the mean. But even so, to me, wrongly uncombined books are "less wrong" than wrongly combined ones, if you see what I mean.

A consequence of uncombined works is that you don't get recommendations. The consequence of wrongly combined works is that you might get recommendations that are off-base. But in this case the wrong combinations are different volumes of the same work so the recommendations for it should be ok