1waltzmn
(Reposted with modifications from the topic "Early Reviewers Redesign and Relaunch.)
This is about the tendency for people to post public reviews that consist of just one or two words, e.g. "Good book." Some people consider this a sufficient review, but most of the books I'm (potentially) interested in are very obscure, often decades old, and almost impossible to find out about. "Good book" tells me nothing that I need to know. So I would like to have some way to get access to better reviews. We now go to the suggestions I made in the wrong place. :-)
It's worth noting that we already have a perfectly good mechanism for saying "good book"/"bad book," and it's the star rating system. If you want to engage in pure black-and-white thinking, and simplify it down to just good/bad, it's possible to give four or two stars, or five or one, as may be. Or you can say good/indifferent/bad, with a 4/3/2 or 5/3/1 rating. Thus "good book" adds nothing to our information that a star rating does not -- and the star rating is subject to statistical examinations which are much harder to apply to two-word reviews.
Furthermore, for a random person to say something is a "good book" conveys no information unless one knows something about that person's taste. I would in general bet that if someone rates a book a "good book" and says nothing more, then I would think it a bad book, but that's just me. :-)
Which point leads me to my feature requests.
Idea #1. The first feature request is that we have some way to filter reviews. A review of "good book" may be useful to the reviewer, but to me, it is a waste of time -- an anti-feature: I have to discard all those useless reviews to see any real reviews.
The problem is that LibraryThing has reviews that the person write for personal use ("I like this book") and reviews that are written for others ("This is a book you should like"). These are completely different functions, but they go in the same Reviews field. Either we need a way to classify reviews based on their function (e.g. personal reviews, scholarly reviews, public non-scholarly reviews) -- or we need a way to filter reviews ("don't show me any reviews less than 50 words"; "Show me only Early Reviewers Reviews"; "only show reviews by people whose reviews average at least 100 words").
Or both, of course. :-)
Idea #2: If I am to make any use at all of the hypothetical "good book" review, I need to know if that person's tastes are anything like mine. In particular, how does this person rate the books that we share? If you and I have say, thirty books which both of us have rated, then we can take a correlation coefficient of those ratings. If that correlation coefficient were high (say 0.8 -- correlation coefficients range from -1 to 1), then I can probably trust your judgment. If the correlation coefficient were low (say -0.7), then I can trust your judgment in reverse: if you say "good book," then I will think it a bad book, and vice versa. If our correlation coefficient is in the middle (-0.2 or +0.15, say), then your taste and mine have nothing to do with each other, and your recommendation is no help to me either way.
It would theoretically be possible to do a "most similar taste" rating for each person on LibraryThing, just as there are "Members with Your Books"; it would also be possible to list one's compatibility with another person's taste for any member whose catalog one visits.
I think this feature would be a tremendous addition to the review system. Even better than getting people to write reviews that actually say something. :-)
This is about the tendency for people to post public reviews that consist of just one or two words, e.g. "Good book." Some people consider this a sufficient review, but most of the books I'm (potentially) interested in are very obscure, often decades old, and almost impossible to find out about. "Good book" tells me nothing that I need to know. So I would like to have some way to get access to better reviews. We now go to the suggestions I made in the wrong place. :-)
It's worth noting that we already have a perfectly good mechanism for saying "good book"/"bad book," and it's the star rating system. If you want to engage in pure black-and-white thinking, and simplify it down to just good/bad, it's possible to give four or two stars, or five or one, as may be. Or you can say good/indifferent/bad, with a 4/3/2 or 5/3/1 rating. Thus "good book" adds nothing to our information that a star rating does not -- and the star rating is subject to statistical examinations which are much harder to apply to two-word reviews.
Furthermore, for a random person to say something is a "good book" conveys no information unless one knows something about that person's taste. I would in general bet that if someone rates a book a "good book" and says nothing more, then I would think it a bad book, but that's just me. :-)
Which point leads me to my feature requests.
Idea #1. The first feature request is that we have some way to filter reviews. A review of "good book" may be useful to the reviewer, but to me, it is a waste of time -- an anti-feature: I have to discard all those useless reviews to see any real reviews.
The problem is that LibraryThing has reviews that the person write for personal use ("I like this book") and reviews that are written for others ("This is a book you should like"). These are completely different functions, but they go in the same Reviews field. Either we need a way to classify reviews based on their function (e.g. personal reviews, scholarly reviews, public non-scholarly reviews) -- or we need a way to filter reviews ("don't show me any reviews less than 50 words"; "Show me only Early Reviewers Reviews"; "only show reviews by people whose reviews average at least 100 words").
Or both, of course. :-)
Idea #2: If I am to make any use at all of the hypothetical "good book" review, I need to know if that person's tastes are anything like mine. In particular, how does this person rate the books that we share? If you and I have say, thirty books which both of us have rated, then we can take a correlation coefficient of those ratings. If that correlation coefficient were high (say 0.8 -- correlation coefficients range from -1 to 1), then I can probably trust your judgment. If the correlation coefficient were low (say -0.7), then I can trust your judgment in reverse: if you say "good book," then I will think it a bad book, and vice versa. If our correlation coefficient is in the middle (-0.2 or +0.15, say), then your taste and mine have nothing to do with each other, and your recommendation is no help to me either way.
It would theoretically be possible to do a "most similar taste" rating for each person on LibraryThing, just as there are "Members with Your Books"; it would also be possible to list one's compatibility with another person's taste for any member whose catalog one visits.
I think this feature would be a tremendous addition to the review system. Even better than getting people to write reviews that actually say something. :-)
2lilithcat
I like your Idea #1, particularly the concept of filtering by number of words, or by people who have an average number. I think function would be difficult, because a) it would require the reviewer to label her review in some way, b) some will have multiple functions, and c) what do you do about the gazillion reviews that are already on LT.
RE: Idea #2 - this works if you don't care about missing out on reviews by people who don't use star ratings, or by people with whom you don't share enough books to make the data helpful.
RE: Idea #2 - this works if you don't care about missing out on reviews by people who don't use star ratings, or by people with whom you don't share enough books to make the data helpful.
3elenchus
Generally have enthusiastic support for this recommendation, and post to voice that support and also to note that I consider it a good thing that the current system allows and even nurtures all of the various uses mentioned in the OP, and likely other uses. It's not a detriment that there are so many different types of reviews, some useful to me and some not; it's a detriment that there aren't tools to make all of those different types more useful to me, and to others.
I'd love to be able to filter reviews by LTers whose collections I've marked as Interesting Libraries, for instance. (That combines Idea #1 and Idea #2.)
I'd love to be able to filter reviews by LTers whose collections I've marked as Interesting Libraries, for instance. (That combines Idea #1 and Idea #2.)
4lilithcat
>3 elenchus:
I'd love to be able to filter reviews by LTers whose collections I've marked as Interesting Libraries
That might actually get me to mark some Interesting Libraries!
I'd love to be able to filter reviews by LTers whose collections I've marked as Interesting Libraries
That might actually get me to mark some Interesting Libraries!
5Bookmarque
I still want to filter reviewers I DON'T want to see. How many years begging? My knee pads are worn out.
6gilroy
So this would deal with someone who has 130 reviews where they all say "Unique" (which I'm thinking might have been an upload error, but still.)
7waltzmn
>2 lilithcat:
I think function would be difficult, because a) it would require the reviewer to label her review in some way, b) some will have multiple functions, and c) what do you do about the gazillion reviews that are already on LT.
All true, and obviously labeling the review would have to be optional. Indeed, most of mine would have multiple functions. Nothing wrong with that; just implement as check boxes or something. And even if most reviews were un-identified, you could still filter on the identifications for the reviews that are tagged. And those who really work hard on their reviews -- me, for instance -- would likely go back and tag the review's function.
RE: Idea #2 - this works if you don't care about missing out on reviews by people who don't use star ratings, or by people with whom you don't share enough books to make the data helpful.
Of course. On the other hand, a person with whom I don't share any books is not likely to share my opinions about books anyway. :-) And note that I was not proposing to use #2 as a filter; I was proposing it as a mechanism for determining whether one is likely to agree with someone else's reviews. If there isn't enough data, then one has to try to evaluate the review on other grounds.
The other thing is, this would encourage me, at least, to put star ratings on more of my books.
Since reviews are voluntary, we can never make a perfect system. The goal is just to try to help. Keep in mind that any information we add, even if incomplete, still makes the database better. I can testify to this from my own work in creating scholarly databases: There were all sorts of features that we didn't originally have. That data will probably never be filled in for every item in the database. But it's still useful for the ones that are added later.
>3 elenchus:
I'd love to be able to filter reviews by LTers whose collections I've marked as Interesting Libraries, for instance. (That combines Idea #1 and Idea #2.)
I like this, too -- because, in most cases, I mark a library as interesting because I have come to trust the other person's reviews. :-) And, just as lilithcat said, it would make me more likely to mark interesting libraries.
One might also want to filter based on most similar library -- though that has its complications, because it's not reciprocal. Of all the libraries on LibraryThing, the one that is closest to mine by weighted average shares just (226/3186) books. I'm just the eighteenth-closest library to his! So his reviews would probably be of more interest to me than the reverse. (In fact, I'm quite sure of that, because I once met the guy. :-)
I note that we have several suggestions for filtering things OUT as well as IN. So it appears, in a perfect world, we want to filter not just on the review type or length but also on the reviewer, and probably on words in the review -- and we want to have a DOES NOT CONTAIN or DOES NOT EQUAL as well as a CONTAINS or EQUALS.
I think function would be difficult, because a) it would require the reviewer to label her review in some way, b) some will have multiple functions, and c) what do you do about the gazillion reviews that are already on LT.
All true, and obviously labeling the review would have to be optional. Indeed, most of mine would have multiple functions. Nothing wrong with that; just implement as check boxes or something. And even if most reviews were un-identified, you could still filter on the identifications for the reviews that are tagged. And those who really work hard on their reviews -- me, for instance -- would likely go back and tag the review's function.
RE: Idea #2 - this works if you don't care about missing out on reviews by people who don't use star ratings, or by people with whom you don't share enough books to make the data helpful.
Of course. On the other hand, a person with whom I don't share any books is not likely to share my opinions about books anyway. :-) And note that I was not proposing to use #2 as a filter; I was proposing it as a mechanism for determining whether one is likely to agree with someone else's reviews. If there isn't enough data, then one has to try to evaluate the review on other grounds.
The other thing is, this would encourage me, at least, to put star ratings on more of my books.
Since reviews are voluntary, we can never make a perfect system. The goal is just to try to help. Keep in mind that any information we add, even if incomplete, still makes the database better. I can testify to this from my own work in creating scholarly databases: There were all sorts of features that we didn't originally have. That data will probably never be filled in for every item in the database. But it's still useful for the ones that are added later.
>3 elenchus:
I'd love to be able to filter reviews by LTers whose collections I've marked as Interesting Libraries, for instance. (That combines Idea #1 and Idea #2.)
I like this, too -- because, in most cases, I mark a library as interesting because I have come to trust the other person's reviews. :-) And, just as lilithcat said, it would make me more likely to mark interesting libraries.
One might also want to filter based on most similar library -- though that has its complications, because it's not reciprocal. Of all the libraries on LibraryThing, the one that is closest to mine by weighted average shares just (226/3186) books. I'm just the eighteenth-closest library to his! So his reviews would probably be of more interest to me than the reverse. (In fact, I'm quite sure of that, because I once met the guy. :-)
I note that we have several suggestions for filtering things OUT as well as IN. So it appears, in a perfect world, we want to filter not just on the review type or length but also on the reviewer, and probably on words in the review -- and we want to have a DOES NOT CONTAIN or DOES NOT EQUAL as well as a CONTAINS or EQUALS.
8AndreasJ
>2 lilithcat:
Years ago, someone said they ignored all reviews that didn’t come with a star rating. I imagine they’d consider missing out reviews from people who don’t use star ratings a positive good.
Years ago, someone said they ignored all reviews that didn’t come with a star rating. I imagine they’d consider missing out reviews from people who don’t use star ratings a positive good.
9paradoxosalpha
>8 AndreasJ: they ignored all reviews that didn’t come with a star rating
That's interesting, and hard for me to understand or sympathize with. The only books I ever rate with stars are ones I have received through LTER. I am honestly uninterested in any mute one-dimensional measure of a book's worth, either from an individual reviewer or aggregated across the LT user base. Since LT offers no explication of what stars mean, they mean almost nothing to me. (I think more stars is "better," but I have anecdotal evidence that people don't always use them that way.)
That's interesting, and hard for me to understand or sympathize with. The only books I ever rate with stars are ones I have received through LTER. I am honestly uninterested in any mute one-dimensional measure of a book's worth, either from an individual reviewer or aggregated across the LT user base. Since LT offers no explication of what stars mean, they mean almost nothing to me. (I think more stars is "better," but I have anecdotal evidence that people don't always use them that way.)
10AndreasJ
>9 paradoxosalpha:
I found it rather baffling too. Unfortunately I don’t recall their reasoning (though I do recall it didn’t ease my bafflement).
I gave up on star ratings pretty quickly because I found it difficult to rate with any consistency.
It follows I wouldn’t be helped by a system to classify reviews by how closely their author’s ratings match mine. But I think it may be useful to have an option to sort reviews by how similar the author’s library is to mine, as per Members With Your Books. Sorting reviews by Friends, Interest Libraries, and Private Watchlist first would also be useful.
I found it rather baffling too. Unfortunately I don’t recall their reasoning (though I do recall it didn’t ease my bafflement).
I gave up on star ratings pretty quickly because I found it difficult to rate with any consistency.
It follows I wouldn’t be helped by a system to classify reviews by how closely their author’s ratings match mine. But I think it may be useful to have an option to sort reviews by how similar the author’s library is to mine, as per Members With Your Books. Sorting reviews by Friends, Interest Libraries, and Private Watchlist first would also be useful.
11Maddz
I've never used star ratings, even for LTER offerings. It's too one-dimensional for my tastes. If I were to use stars, I'd want one rating for (at least):
Storyline / plot * (banal)
Worldbuilding **** (interesting)
Writing **** (well-written)
Editing ** (poor - discontinuity errors abound)
Did I like it or not? *** (OK - I'd read more from this author but I wouldn't bother seeking any out)
So I might loathe the story (or it was just plain banal), but the worldbuilding was interesting, it was well-written but poorly edited, and I liked it, which would give 14/25 stars or a little under 3* overall.
Ideally, we'd want user defined stars, perhaps with a notes area for each star to briefly explain your rating.
Storyline / plot * (banal)
Worldbuilding **** (interesting)
Writing **** (well-written)
Editing ** (poor - discontinuity errors abound)
Did I like it or not? *** (OK - I'd read more from this author but I wouldn't bother seeking any out)
So I might loathe the story (or it was just plain banal), but the worldbuilding was interesting, it was well-written but poorly edited, and I liked it, which would give 14/25 stars or a little under 3* overall.
Ideally, we'd want user defined stars, perhaps with a notes area for each star to briefly explain your rating.
12igorken
>11 Maddz: I do use ratings to indicate how much I enjoyed/appreciated something, and you are absolutely right that it is very difficult to remain consistent. Writing a proper review unfortunately takes too much time.
An interface that easily (click, click, click) allows me to rate multiple parameters, would be pretty neat. Perhaps I may actually have a go at using the review field to do that (something to similar to what you posted above) to see how that goes.
An interface that easily (click, click, click) allows me to rate multiple parameters, would be pretty neat. Perhaps I may actually have a go at using the review field to do that (something to similar to what you posted above) to see how that goes.
13Caramellunacy
>3 elenchus: I agree that I would be very interested in being able to filter / sort to top reviews by my "connections" - interesting libraries or friends or similar (similar to the GR feature) as I have already determined that their opinions/taste correlate.
Similarly, I would like to be able to "mute" certain reviewers (which would be of more use to me than filtering by type/keyword, though I can certainly see how that would be useful for others).
Similarly, I would like to be able to "mute" certain reviewers (which would be of more use to me than filtering by type/keyword, though I can certainly see how that would be useful for others).
14lilithcat
>10 AndreasJ:
I gave up on star ratings pretty quickly because I found it difficult to rate with any consistency.
They also mean different things to different people. I've seen reviews that were entirely positive, no negatives mentioned, but the book is given 3 stars, and vice versa.
I gave up on star ratings pretty quickly because I found it difficult to rate with any consistency.
They also mean different things to different people. I've seen reviews that were entirely positive, no negatives mentioned, but the book is given 3 stars, and vice versa.
15booksaplenty1949
I wish there was a way to filter out the reviews that are actually notes the member has written to him/herself about the book’s condition or its location in the house. Sometimes the former are so detailed I suspect the book is for sale.
16booksaplenty1949
Then there are the “reviews” like “Read this in high school; don’t remember anything about it.” or variations of the line about War and Peace—-“It concerns Russia.” By definition a real review is evaluative.
17waltzmn
>10 AndreasJ:
I gave up on star ratings pretty quickly because I found it difficult to rate with any consistency.
For those of you who didn't see the "back story" on these ideas, which is in the thread about the new Early Reviewers, realize that this came out of a person who considered himself a reviewer but felt that "Good book" or "Bad book" was all the review he would ever need to give. (I'm pretty sure it was a him, because women are rarely that obnoxious. :-)
Compared to that, the star rating system is deep and rich. :-)
Obviously, as with everything else, we can't expect everyone to award star ratings. And everyone's will be different. (I, for instance, work pretty hard to keep my average rating very close to 3.0. The flip side is, I don't give star ratings all that often -- mostly I give them when I write a review, as a summary of my overall opinion.) One could always include ratings based on multiple categories, such as Maddz proposes, in the body of one's review -- and then, perhaps, use the star rating as an average.
But note that I'm not really arguing for star ratings as a rating system. The fact that a book gets a high star rating means very little to me, because my tastes are very different from the general public. The proposal I made was for a way to see if a particular reviewer's tastes seem to match mine, so that I can know how far to trust that reviewer's reviews. Such a rating system must be based on some data. Basing it on star ratings has the advantage of being 100% repeatable and relatively simple, algorithmically. Will it always work? Of course not. Is there anything that is simultaneously at least as easy and at least as effective? I very, very much doubt it, particularly within the data set that LibraryThing already collects. (This proposal would not require any changes to the database; just some new code. Changing code is a lot safer than changing the underlying database!)
I gave up on star ratings pretty quickly because I found it difficult to rate with any consistency.
For those of you who didn't see the "back story" on these ideas, which is in the thread about the new Early Reviewers, realize that this came out of a person who considered himself a reviewer but felt that "Good book" or "Bad book" was all the review he would ever need to give. (I'm pretty sure it was a him, because women are rarely that obnoxious. :-)
Compared to that, the star rating system is deep and rich. :-)
Obviously, as with everything else, we can't expect everyone to award star ratings. And everyone's will be different. (I, for instance, work pretty hard to keep my average rating very close to 3.0. The flip side is, I don't give star ratings all that often -- mostly I give them when I write a review, as a summary of my overall opinion.) One could always include ratings based on multiple categories, such as Maddz proposes, in the body of one's review -- and then, perhaps, use the star rating as an average.
But note that I'm not really arguing for star ratings as a rating system. The fact that a book gets a high star rating means very little to me, because my tastes are very different from the general public. The proposal I made was for a way to see if a particular reviewer's tastes seem to match mine, so that I can know how far to trust that reviewer's reviews. Such a rating system must be based on some data. Basing it on star ratings has the advantage of being 100% repeatable and relatively simple, algorithmically. Will it always work? Of course not. Is there anything that is simultaneously at least as easy and at least as effective? I very, very much doubt it, particularly within the data set that LibraryThing already collects. (This proposal would not require any changes to the database; just some new code. Changing code is a lot safer than changing the underlying database!)
18aspirit
>17 waltzmn: a note about the inspiration: I'm fairly certain everyone who posted in the New Features - EarlyReviewers thread typically writes much longer, detailed reviews than LibraryThing's bare minimum, and nothing I saw in multiple reads of the page suggested decreasing the word count of the ER minimum.
>1 waltzmn: "This is about the tendency for people to post public reviews that consist of just one or two words, e.g. 'Good book.'" Despite looking at new-to-me works (many which are obscure, older than the typical college student, and difficult to find) just about every day, I rarely see these reviews.
Maybe the extremely short reviews are more common on a specific types of books?If so, I think identifying which types could help the LT team in the development of Idea #1. On second thought, I suspect focusing on review types such as "personal" or "semi-professional" to identify purpose might garner better attention from the team. The tiniest reviews are easy to miss while browsing.
But also, one of the fastest ways to find one- and two-word reviews (that should not be flagged) is to look at the site owner's reviews (a mix of styles, like many non-professional reviewers have). Members trying to judge other members (what Idea #2 vaguely looks like coming after criticism as it does) instead of focusing on the books hasn't seemed to have gone over well with him in the past.
>1 waltzmn: "This is about the tendency for people to post public reviews that consist of just one or two words, e.g. 'Good book.'" Despite looking at new-to-me works (many which are obscure, older than the typical college student, and difficult to find) just about every day, I rarely see these reviews.
Maybe the extremely short reviews are more common on a specific types of books?
But also, one of the fastest ways to find one- and two-word reviews (that should not be flagged) is to look at the site owner's reviews (a mix of styles, like many non-professional reviewers have). Members trying to judge other members (what Idea #2 vaguely looks like coming after criticism as it does) instead of focusing on the books hasn't seemed to have gone over well with him in the past.
19waltzmn
>18 aspirit:
I'm fairly certain everyone who posted in the New Features - EarlyReviewers thread typically writes much longer, detailed reviews than LibraryThing's bare minimum.
I think that is true with one exception. The one exception is a series of posts by medwards429, which have been deleted by the author. These posts proposed "Good book" as an acceptable review. It is a review that meets the requirements, but I don't consider it useful.
I'm also interested in the first line of this new feature request. "This is about the tendency for people to post public reviews that consist of just one or two words, e.g. 'Good book.'" Despite looking at new-to-me works (many which are obscure, older than the typical college student, and difficult to find) just about every day, I rarely see these reviews. Maybe it's more common on a specific types of books? If so, I think identifying which types could help the LT team in the development of Idea #1.
I think your answer lies in the statement "Despite looking at new-to-me works (many which are obscure, older than the typical college student, and difficult to find)" :-) I agree that such books, if reviewed at all, are much more likely to get substantial reviews. But it's not guaranteed.
And while my library is exceptionally obscure :-), I encounter short reviews sometimes. Right now, the #3 review down my new reviews feed is a "review" by JFB87 of Erik Larson's Dead Wake: The Last Crossing of the Lusitania. It consists of five stars and the single word "Unique." Maybe it's a five-star book if you ignore a number of sloppy errors (for some reason, books about the Lusitania seem to be pretty bad; Larson's isn't the worst, but doesn't he have a factchecker?), but "Unique" is certainly no more helpful a review than was medwards429's "Good book."
#5 in my feed is William Ryan's Noah's Flood, reviewed by capewood. This is fuller, but still only about 43 words, and three of them are not relevant to a review, 38 are a brief summary of the content, and the only assessment is "interesting read."
Before anyone rags on me about having Harry Potter in my library, I'll note that I've written a scholarly book on Harry Potter folklore (Harry Potter's Folklore World: On Myth and Magic), so I have reasons to own the books. #13 down my feed is Boreque's review of Harry Potter and the Deathly Hallows. The review says "Great book, nice ending."
#18 in my list is David Ewen's Home Book of Musical Knowledge. The review by DavidDuBois reads "ENTERTAINMENT CENTER."
I suspect we could establish a correlation between the popularity of the book and the shortness of the review. :-)
Still, even in my extremely obscure library, I'm getting a lot of useless reviews. Of the latest 30 reviews in my feed as of this writing, if I were coming at the book cold, I would consider only nine to be long enough to be useful (that's not saying I agree with them -- in three cases, I don't).
So my first personal first desire is to drop any review less than about 100 words. That's straightforward. It would also be nice to block people like "DavidDuBois" who misuse the reviews field. Ideally, I'd also like to drop the reviews by people who don't care that books are full of errors, but that's admittedly harder. :-) My star rating correlation would not solve that, but it might help.
I'm fairly certain everyone who posted in the New Features - EarlyReviewers thread typically writes much longer, detailed reviews than LibraryThing's bare minimum.
I think that is true with one exception. The one exception is a series of posts by medwards429, which have been deleted by the author. These posts proposed "Good book" as an acceptable review. It is a review that meets the requirements, but I don't consider it useful.
I'm also interested in the first line of this new feature request. "This is about the tendency for people to post public reviews that consist of just one or two words, e.g. 'Good book.'" Despite looking at new-to-me works (many which are obscure, older than the typical college student, and difficult to find) just about every day, I rarely see these reviews. Maybe it's more common on a specific types of books? If so, I think identifying which types could help the LT team in the development of Idea #1.
I think your answer lies in the statement "Despite looking at new-to-me works (many which are obscure, older than the typical college student, and difficult to find)" :-) I agree that such books, if reviewed at all, are much more likely to get substantial reviews. But it's not guaranteed.
And while my library is exceptionally obscure :-), I encounter short reviews sometimes. Right now, the #3 review down my new reviews feed is a "review" by JFB87 of Erik Larson's Dead Wake: The Last Crossing of the Lusitania. It consists of five stars and the single word "Unique." Maybe it's a five-star book if you ignore a number of sloppy errors (for some reason, books about the Lusitania seem to be pretty bad; Larson's isn't the worst, but doesn't he have a factchecker?), but "Unique" is certainly no more helpful a review than was medwards429's "Good book."
#5 in my feed is William Ryan's Noah's Flood, reviewed by capewood. This is fuller, but still only about 43 words, and three of them are not relevant to a review, 38 are a brief summary of the content, and the only assessment is "interesting read."
Before anyone rags on me about having Harry Potter in my library, I'll note that I've written a scholarly book on Harry Potter folklore (Harry Potter's Folklore World: On Myth and Magic), so I have reasons to own the books. #13 down my feed is Boreque's review of Harry Potter and the Deathly Hallows. The review says "Great book, nice ending."
#18 in my list is David Ewen's Home Book of Musical Knowledge. The review by DavidDuBois reads "ENTERTAINMENT CENTER."
I suspect we could establish a correlation between the popularity of the book and the shortness of the review. :-)
Still, even in my extremely obscure library, I'm getting a lot of useless reviews. Of the latest 30 reviews in my feed as of this writing, if I were coming at the book cold, I would consider only nine to be long enough to be useful (that's not saying I agree with them -- in three cases, I don't).
So my first personal first desire is to drop any review less than about 100 words. That's straightforward. It would also be nice to block people like "DavidDuBois" who misuse the reviews field. Ideally, I'd also like to drop the reviews by people who don't care that books are full of errors, but that's admittedly harder. :-) My star rating correlation would not solve that, but it might help.
20aspirit
Note: I revised my post above after my spouse said my original words sounded sarcastic, which wasn't my intention. Then I got to thinking more about what was bothering me enough to come across that way.
I'll be back to read >18 aspirit: after taking a household situation. Thanks for understanding.
I'll be back to read >18 aspirit: after taking a household situation. Thanks for understanding.
21lilithcat
>19 waltzmn:
The one exception is a series of posts by @medwards429, which have been deleted by the author. These posts proposed "Good book" as an acceptable review.
Quite the contrary! While he did note that LT accepted such reviews, he himself went on and on about what good reviews ought to be (primarily "do it my way", which, if you look at his reviews, are definitely not of the "good book" variety). You can see his opinion in this reply to one of his posts: https://www.librarything.com/topic/339404#7758018 His posts are there described as "off-topic criticism of readers who leave short reviews (acceptable by the standards of this site)"
Right now, the #3 review down my new reviews feed is a "review" by @JFB87 of Erik Larson's Dead Wake: The Last Crossing of the Lusitania. It consists of five stars and the single word "Unique."
That may be an import error, as that is the complete content of all that person's reviews.
Before anyone rags on me about having Harry Potter in my library
Why would anyone do that?
The one exception is a series of posts by @medwards429, which have been deleted by the author. These posts proposed "Good book" as an acceptable review.
Quite the contrary! While he did note that LT accepted such reviews, he himself went on and on about what good reviews ought to be (primarily "do it my way", which, if you look at his reviews, are definitely not of the "good book" variety). You can see his opinion in this reply to one of his posts: https://www.librarything.com/topic/339404#7758018 His posts are there described as "off-topic criticism of readers who leave short reviews (acceptable by the standards of this site)"
Right now, the #3 review down my new reviews feed is a "review" by @JFB87 of Erik Larson's Dead Wake: The Last Crossing of the Lusitania. It consists of five stars and the single word "Unique."
That may be an import error, as that is the complete content of all that person's reviews.
Before anyone rags on me about having Harry Potter in my library
Why would anyone do that?
22waltzmn
>21 lilithcat:
The one exception is a series of posts by medwards429, which have been deleted by the author. These posts proposed "Good book" as an acceptable review.
Quite the contrary! While he did note that LT accepted such reviews, he himself went on and on about what good reviews ought to be (primarily "do it my way", which, if you look at his reviews, are definitely not of the "good book" variety).
OK, my apologies. I skimmed part of the thread and took the wrong impression. :-( And couldn't check back because it was deleted. I didn't think to check the other messages that quoted it.
Right now, the #3 review down my new reviews feed is a "review" by JFB87 of Erik Larson's Dead Wake: The Last Crossing of the Lusitania. It consists of five stars and the single word "Unique."
That may be an import error, as that is the complete content of all that person's reviews.
Possibly, but not really relevant, because the review is still there. And a person just looking at the review of a particular book will not know that.
Before anyone rags on me about having Harry Potter in my library
Why would anyone do that?
Here I have to blather. :-) Apologies.
The reason people harass me about it is because it's popular. :-) Also, it's not a good fit for my library on other grounds -- by my rough count, my library is over 80% non-fiction (LibraryThing says 75%, but it doesn't know what a lot of my books are, they're so obscure. :-) Also, I believe that a lot of my critical editions of things like Shakespeare and Chaucer are counted as fiction, when what I use them for is the apparatus (non-fiction), not the text.
The obscurity of my library is sort of the point: I don't need reviews of Harry Potter; there are plenty of those. But many of the books I'm trying to find out about have only one or two reviews anywhere. So finding a review I can trust is very important to me. Bogus reviews like "Unique" have not really affected my book evaluation much, but they make me think about how they could affect my purchases. There have been a few instances, e.g., where fiction books (historical novels) have been presented as -- even reviewed as -- non-fiction. Historical novels are no use to me. So those reviews can cause me difficulty.
The one exception is a series of posts by medwards429, which have been deleted by the author. These posts proposed "Good book" as an acceptable review.
Quite the contrary! While he did note that LT accepted such reviews, he himself went on and on about what good reviews ought to be (primarily "do it my way", which, if you look at his reviews, are definitely not of the "good book" variety).
OK, my apologies. I skimmed part of the thread and took the wrong impression. :-( And couldn't check back because it was deleted. I didn't think to check the other messages that quoted it.
Right now, the #3 review down my new reviews feed is a "review" by JFB87 of Erik Larson's Dead Wake: The Last Crossing of the Lusitania. It consists of five stars and the single word "Unique."
That may be an import error, as that is the complete content of all that person's reviews.
Possibly, but not really relevant, because the review is still there. And a person just looking at the review of a particular book will not know that.
Before anyone rags on me about having Harry Potter in my library
Why would anyone do that?
Here I have to blather. :-) Apologies.
The reason people harass me about it is because it's popular. :-) Also, it's not a good fit for my library on other grounds -- by my rough count, my library is over 80% non-fiction (LibraryThing says 75%, but it doesn't know what a lot of my books are, they're so obscure. :-) Also, I believe that a lot of my critical editions of things like Shakespeare and Chaucer are counted as fiction, when what I use them for is the apparatus (non-fiction), not the text.
The obscurity of my library is sort of the point: I don't need reviews of Harry Potter; there are plenty of those. But many of the books I'm trying to find out about have only one or two reviews anywhere. So finding a review I can trust is very important to me. Bogus reviews like "Unique" have not really affected my book evaluation much, but they make me think about how they could affect my purchases. There have been a few instances, e.g., where fiction books (historical novels) have been presented as -- even reviewed as -- non-fiction. Historical novels are no use to me. So those reviews can cause me difficulty.
23AndreasJ
>17 waltzmn:
If I’d tried to simply divide books into good and bad, rather than use a ten point scale, I’d surely found it easier to be consistent. On the other hand, it’s hard to see how it’d be much use to myself or anyone else - I exercise enough due diligence before getting a book that the large majority of ones I end up reading would be classed as good.
If I’d tried to simply divide books into good and bad, rather than use a ten point scale, I’d surely found it easier to be consistent. On the other hand, it’s hard to see how it’d be much use to myself or anyone else - I exercise enough due diligence before getting a book that the large majority of ones I end up reading would be classed as good.
24paradoxosalpha
>23 AndreasJ:
Large swathes of the internet seem to treat 5 stars as "good" and less than 5 as "bad."
Large swathes of the internet seem to treat 5 stars as "good" and less than 5 as "bad."
25AndreasJ
>24 paradoxosalpha:
That large swathes of the Internet are of little use to me or anyone else is not, I think, a very surprising conclusion.
That large swathes of the Internet are of little use to me or anyone else is not, I think, a very surprising conclusion.
26waltzmn
>23 AndreasJ:
If I’d tried to simply divide books into good and bad, rather than use a ten point scale, I’d surely found it easier to be consistent. On the other hand, it’s hard to see how it’d be much use to myself or anyone else - I exercise enough due diligence before getting a book that the large majority of ones I end up reading would be classed as good.
This isn't really relevant to the feature requests, since many people do use star ratings, but I'll give some thoughts.
I'll grant the point on consistency -- if you only have two choices, books will pretty definitely fall into one category or the other. On the other hand, I would personally find it very confining to only be able to rate "good" or "bad." There are "great" books, and there are books which are "good" but not "great." There are "bad" books, and there are "terrible" books. I agree that, for any given "good" book, I might rate it 3.5 stars one week and 4 the next -- but I wouldn't rate it 3.0 or 5.0. I might rate a great book 4.5 or 5.0, but not 4.0 or lower. And so forth.
The other thing is that, for me at least, there are a lot of complications here. As an example, or maybe a thought experiment, take the infamous King Richard III. I am a "revisionist." That is, I think that Richard wasn't as black as painted -- certainly not as bad as in Shakespeare's The Tragedy of Richard III or Alison Weir's black-fantasy-she-pretends-is-history The Princes in the Tower.
But of my books on the topic, I much prefer (say) Charles Derek Ross's Richard III to Paul Murray Kendall's book of the same title, even though Ross is a traditionalist and Kendall a revisionist; Ross is simply a better scholar. It's just that I prefer Anthony Cheetham's The Life and Times of Richard III to either, because it's revisionist, but less absurdly revisionist than Kendall's work and is more scholarly.
So I even if I can't reliably give star ratings, I can reliably set an order: Cheetham is better than Ross, which is better than Kendall, which is better than Weir. From there, how hard is it to go to a star system? Not hard: Cheetham gets four stars (there is no five-star book on Richard III that I know of), Ross gets three, Kendall two, Weir one. Or maybe Cheetham four, Ross three and a half, Kendall two and a half, Weir a half. (Weir really ought to get a negative score, because she doesn't call her work fiction!) That's not too much error.
So you could probably set up a reference system.
Of course, it only matters if you're trying to find people who view books the same way you do. If you aren't, then it's minor.
If I’d tried to simply divide books into good and bad, rather than use a ten point scale, I’d surely found it easier to be consistent. On the other hand, it’s hard to see how it’d be much use to myself or anyone else - I exercise enough due diligence before getting a book that the large majority of ones I end up reading would be classed as good.
This isn't really relevant to the feature requests, since many people do use star ratings, but I'll give some thoughts.
I'll grant the point on consistency -- if you only have two choices, books will pretty definitely fall into one category or the other. On the other hand, I would personally find it very confining to only be able to rate "good" or "bad." There are "great" books, and there are books which are "good" but not "great." There are "bad" books, and there are "terrible" books. I agree that, for any given "good" book, I might rate it 3.5 stars one week and 4 the next -- but I wouldn't rate it 3.0 or 5.0. I might rate a great book 4.5 or 5.0, but not 4.0 or lower. And so forth.
The other thing is that, for me at least, there are a lot of complications here. As an example, or maybe a thought experiment, take the infamous King Richard III. I am a "revisionist." That is, I think that Richard wasn't as black as painted -- certainly not as bad as in Shakespeare's The Tragedy of Richard III or Alison Weir's black-fantasy-she-pretends-is-history The Princes in the Tower.
But of my books on the topic, I much prefer (say) Charles Derek Ross's Richard III to Paul Murray Kendall's book of the same title, even though Ross is a traditionalist and Kendall a revisionist; Ross is simply a better scholar. It's just that I prefer Anthony Cheetham's The Life and Times of Richard III to either, because it's revisionist, but less absurdly revisionist than Kendall's work and is more scholarly.
So I even if I can't reliably give star ratings, I can reliably set an order: Cheetham is better than Ross, which is better than Kendall, which is better than Weir. From there, how hard is it to go to a star system? Not hard: Cheetham gets four stars (there is no five-star book on Richard III that I know of), Ross gets three, Kendall two, Weir one. Or maybe Cheetham four, Ross three and a half, Kendall two and a half, Weir a half. (Weir really ought to get a negative score, because she doesn't call her work fiction!) That's not too much error.
So you could probably set up a reference system.
Of course, it only matters if you're trying to find people who view books the same way you do. If you aren't, then it's minor.
27bokai
This is an interesting discussion to have and is probably pointing to a problem that is a lot more difficult to solve than it seems at first glance.
It's really hard to systematize subjective information. And there are two layers of this here. There are questions of the subjective conclusions a review makes towards a book, and the subjective value a given review may have to someone else reading it.
We already have some minor attempt a dealing with the second issue with the thumbs up votes, which you can already sort by. It seems this hasn't been mentioned in the thread yet so I'd like to put it out there.
Because there are already ways to sort reviews, perhaps adding a sort by length would be an easier function to implement than a hard filter.
The similar taste rating I can see as being a fun new feature. It would fit well in sort of "fun with data" niche that librarything has and that I enjoy so much (other people hang out on the zeitgeist page, right???)
>4 lilithcat:
There is actually already a way to see reviews posted by your connections. If you go into the Home > Connections tab you can see recent activity by friends and interesting libraries, and filter by reviews. I actually used to use this often many years ago, but am now just remembering it exists from this discussion. The does not however let you go to a particular work and see what reviews your connections may have posted from there. This may be something simple to implement as a "connections on top" sort.
It's really hard to systematize subjective information. And there are two layers of this here. There are questions of the subjective conclusions a review makes towards a book, and the subjective value a given review may have to someone else reading it.
We already have some minor attempt a dealing with the second issue with the thumbs up votes, which you can already sort by. It seems this hasn't been mentioned in the thread yet so I'd like to put it out there.
Because there are already ways to sort reviews, perhaps adding a sort by length would be an easier function to implement than a hard filter.
The similar taste rating I can see as being a fun new feature. It would fit well in sort of "fun with data" niche that librarything has and that I enjoy so much (other people hang out on the zeitgeist page, right???)
>4 lilithcat:
There is actually already a way to see reviews posted by your connections. If you go into the Home > Connections tab you can see recent activity by friends and interesting libraries, and filter by reviews. I actually used to use this often many years ago, but am now just remembering it exists from this discussion. The does not however let you go to a particular work and see what reviews your connections may have posted from there. This may be something simple to implement as a "connections on top" sort.
28paradoxosalpha
>27 bokai:
Excellent points, all.
Excellent points, all.
29waltzmn
>27 bokai:
We already have some minor attempt a dealing with the second issue with the thumbs up votes, which you can already sort by. It seems this hasn't been mentioned in the thread yet so I'd like to put it out there.
This is a good point, but high rankings for a review don't automatically mean that one will agree with it. I'll give a deliberately divisive example :-) -- C. S. Lewis's The Lion, the Witch, and the Wardrobe. This is, of course, an extremely popular work. And it bugs me a lot. First there is J. R. R. Tolkien's objection, that it is a mythological hotch-potch, which irritates me quite a bit. Then there is the Christian heresy in that is obvious to me but that no one else seems to see. (Romans 6:10 -- "The death he died, he died to sin, once for all." For Aslan to die again, to save Edmund, is to destroy the whole purpose of the Incarnation.)
It doesn't matter whether you agree with me or not -- obviously most don't, or they wouldn't keep putting out C. S. Lewis products to the point where I wonder about idolatry. :-) The fact that I'm an autistic nitpicker is no doubt relevant here :-) -- internal inconsistency drives me nuts. But the point is that the most-upvoted review about that book is one that I disagree with intensely. Being popular doesn't mean that a review would be to my taste.
Because there are already ways to sort reviews, perhaps adding a sort by length would be an easier function to implement than a hard filter.
That would be a nice start, and as you say, it should be easy. Maybe that could be a short-term idea and a better set of filters a long-term goal. Or even three stages: 1. Sort by length. 2. True filtering. 3. Alter the database to try to encourage people to mage reviews more easily categorized.
Doing the "reviews-alike" feature would be harder than (1), possibly harder than (2), probably easier than (3). (Based on my own coding experience, not based on any knowledge of LibraryThing's back end. I could be wrong.)
We already have some minor attempt a dealing with the second issue with the thumbs up votes, which you can already sort by. It seems this hasn't been mentioned in the thread yet so I'd like to put it out there.
This is a good point, but high rankings for a review don't automatically mean that one will agree with it. I'll give a deliberately divisive example :-) -- C. S. Lewis's The Lion, the Witch, and the Wardrobe. This is, of course, an extremely popular work. And it bugs me a lot. First there is J. R. R. Tolkien's objection, that it is a mythological hotch-potch, which irritates me quite a bit. Then there is the Christian heresy in that is obvious to me but that no one else seems to see. (Romans 6:10 -- "The death he died, he died to sin, once for all." For Aslan to die again, to save Edmund, is to destroy the whole purpose of the Incarnation.)
It doesn't matter whether you agree with me or not -- obviously most don't, or they wouldn't keep putting out C. S. Lewis products to the point where I wonder about idolatry. :-) The fact that I'm an autistic nitpicker is no doubt relevant here :-) -- internal inconsistency drives me nuts. But the point is that the most-upvoted review about that book is one that I disagree with intensely. Being popular doesn't mean that a review would be to my taste.
Because there are already ways to sort reviews, perhaps adding a sort by length would be an easier function to implement than a hard filter.
That would be a nice start, and as you say, it should be easy. Maybe that could be a short-term idea and a better set of filters a long-term goal. Or even three stages: 1. Sort by length. 2. True filtering. 3. Alter the database to try to encourage people to mage reviews more easily categorized.
Doing the "reviews-alike" feature would be harder than (1), possibly harder than (2), probably easier than (3). (Based on my own coding experience, not based on any knowledge of LibraryThing's back end. I could be wrong.)
30paradoxosalpha
>29 waltzmn: ...one that I disagree with intensely. Being popular doesn't mean that a review would be to my taste.
I enjoy reading reviews that I disagree with, if they are thoughtful and well-written!
I enjoy reading reviews that I disagree with, if they are thoughtful and well-written!
31Stevil2001
Yes, I would never think that "surfacing good reviews" means "surfacing ones I agree with."
32bokai
>29 waltzmn:
You're right that the thumbs up has a number of issues. If you take other places where such sorting systems are in place, like the Steam marketplace, community endorsement rewards entertaining or even joke comments more than they tend to reward deep, lengthy, or nuanced takes. So prioritizing community points can sometimes incentivize the wrong sort of content. Steam eventually had to add a "funny" endorsement separately to try to fix this.
If we are trying to get libarything to show us only the reviews that we like best and agree with, IMO that is nearly an impossible challenge. As human beings we're simply not consistent enough for an algorithm to predict our responses to creative output (Netflix and Amazon may try, but these sorts of algorithms often hide less popular material as they illuminate things "we might like" which is not a tradeoff I personally want). That's not to say that we can't improve things, only that the challenge is quite high for any solution in which a majority of users are satisfied. We can probably make broad improvements, but many exceptions will slip through.
The C. S. Lewis example is an interesting test. What sort of filters can we imagine where the system would know you don't want to see that review? Is it just a high star score for a book you may score low? Does this mean you need to score a book before the function will be useful? Will this help when trying to figure out if you want to read a book you haven't scored yet? Will comparative aggregate scores between users tell you if your interests truly align? Or will maybe there not be enough agreement between users for that to be the case? My friends and I have generally similar taste in media, but we still will find ourselves arguing about whether or not something we all just watched or read was good or not.
You're right that the thumbs up has a number of issues. If you take other places where such sorting systems are in place, like the Steam marketplace, community endorsement rewards entertaining or even joke comments more than they tend to reward deep, lengthy, or nuanced takes. So prioritizing community points can sometimes incentivize the wrong sort of content. Steam eventually had to add a "funny" endorsement separately to try to fix this.
If we are trying to get libarything to show us only the reviews that we like best and agree with, IMO that is nearly an impossible challenge. As human beings we're simply not consistent enough for an algorithm to predict our responses to creative output (Netflix and Amazon may try, but these sorts of algorithms often hide less popular material as they illuminate things "we might like" which is not a tradeoff I personally want). That's not to say that we can't improve things, only that the challenge is quite high for any solution in which a majority of users are satisfied. We can probably make broad improvements, but many exceptions will slip through.
The C. S. Lewis example is an interesting test. What sort of filters can we imagine where the system would know you don't want to see that review? Is it just a high star score for a book you may score low? Does this mean you need to score a book before the function will be useful? Will this help when trying to figure out if you want to read a book you haven't scored yet? Will comparative aggregate scores between users tell you if your interests truly align? Or will maybe there not be enough agreement between users for that to be the case? My friends and I have generally similar taste in media, but we still will find ourselves arguing about whether or not something we all just watched or read was good or not.
33waltzmn
>30 paradoxosalpha:
I enjoy reading reviews that I disagree with, if they are thoughtful and well-written!
You are of course right, and I shouldn't have said "disagree with"; I should have said something like "are concerned with the same issues I'm concerned with." On the other hand, why do we read reviews? Occasionally it is to see what others think about a book we have already read -- but I, at least, usually read a review to find out if I want to buy the book. So I can't know if I disagree with the review. :-) The point of my C. S. Lewis example is that if I had bought The Lion, the Witch, and the Wardrobe based on that review, I would have been intensely disappointed by the outcome and disagreed with the review too late for it to do me any good.
>32 bokai:
The C. S. Lewis example is an interesting test. What sort of filters can we imagine where the system would know you don't want to see that review? Is it just a high star score for a book you may score low? Does this mean you need to score a book before the function will be useful? Will this help when trying to figure out if you want to read a book you haven't scored yet? Will comparative aggregate scores between users tell you if your interests truly align? Or will maybe there not be enough agreement between users for that to be the case? My friends and I have generally similar taste in media, but we still will find ourselves arguing about whether or not something we all just watched or read was good or not.
This is indeed the question. It's why I proposed my matching-book-rankings idea.
Other things that might work to some extent might be:
-- Similar libraries (if the reviewer's library is similar to yours, then the review rates higher). On the other hand, this assumes one has similar libraries. The evidence is that I, frankly, don't. :-)
-- Interesting libraries. This probably means that you've read a lot of the same books as the owner of the other library, or have approved of a lot of their reviews, and so have similar tastes
-- Similar library classifications. You could perhaps do a correlation coefficient with the Dewey Decimal classifications (you and I might not have a lot of the same books, but if both of our libraries are 45% history, we probably have somewhat similar tastes)
-- Similarly with Genres, although I don't trust that one much because the genre list doesn't match my personal library
Possibly one could make some sort of weighted average out of these. But I suspect that would be difficult and slow to calculate. The correlation of book ratings would probably be easier.
I enjoy reading reviews that I disagree with, if they are thoughtful and well-written!
You are of course right, and I shouldn't have said "disagree with"; I should have said something like "are concerned with the same issues I'm concerned with." On the other hand, why do we read reviews? Occasionally it is to see what others think about a book we have already read -- but I, at least, usually read a review to find out if I want to buy the book. So I can't know if I disagree with the review. :-) The point of my C. S. Lewis example is that if I had bought The Lion, the Witch, and the Wardrobe based on that review, I would have been intensely disappointed by the outcome and disagreed with the review too late for it to do me any good.
>32 bokai:
The C. S. Lewis example is an interesting test. What sort of filters can we imagine where the system would know you don't want to see that review? Is it just a high star score for a book you may score low? Does this mean you need to score a book before the function will be useful? Will this help when trying to figure out if you want to read a book you haven't scored yet? Will comparative aggregate scores between users tell you if your interests truly align? Or will maybe there not be enough agreement between users for that to be the case? My friends and I have generally similar taste in media, but we still will find ourselves arguing about whether or not something we all just watched or read was good or not.
This is indeed the question. It's why I proposed my matching-book-rankings idea.
Other things that might work to some extent might be:
-- Similar libraries (if the reviewer's library is similar to yours, then the review rates higher). On the other hand, this assumes one has similar libraries. The evidence is that I, frankly, don't. :-)
-- Interesting libraries. This probably means that you've read a lot of the same books as the owner of the other library, or have approved of a lot of their reviews, and so have similar tastes
-- Similar library classifications. You could perhaps do a correlation coefficient with the Dewey Decimal classifications (you and I might not have a lot of the same books, but if both of our libraries are 45% history, we probably have somewhat similar tastes)
-- Similarly with Genres, although I don't trust that one much because the genre list doesn't match my personal library
Possibly one could make some sort of weighted average out of these. But I suspect that would be difficult and slow to calculate. The correlation of book ratings would probably be easier.
34AndreasJ
>33 waltzmn:
Your median number of members who share your books is 45, so I don't think your library is too obscure for similar libraries to be useful. It pretty reliably finds me libraries of interest, and my median is 30.
Your median number of members who share your books is 45, so I don't think your library is too obscure for similar libraries to be useful. It pretty reliably finds me libraries of interest, and my median is 30.
35waltzmn
>34 AndreasJ:
Your median number of members who share your books is 45, so I don't think your library is too obscure for similar libraries to be useful. It pretty reliably finds me libraries of interest, and my median is 30.
I hadn't seen that statistic. Interesting. I'll have to find that statistic. :-)
But I've looked at my similar libraries. The first one on my list is moderately interesting (and in fact is someone whom I have met, decades ago), but the next few are not.
The real reason for my statement is that I looked at the library most like mine (https://www.librarything.com/profile/dankeding). His library is the closest to mine -- but, as of the last time I looked, I was only the seventeenth-closest to his. It's long branch assimilation -- my library is so weird that a library only moderately like mine comes up closest.
Possibly the high median score is because I have a lot of books that a lot of others have (many editions of The Iliad, The Odyssey, The Complete Works of Shakespeare, sundry Bible editions, a few popular fantasy works and common references). But these are not the interesting parts of my library. The interesting parts are things like Traditional singers and songs from Ontario (three other copies), The bothy songs and ballads of Aberdeen, Banff and Moray, Angus and the Mearns (three other copies), Folksongs of Mississippi and their background (one other copy).
Still, I didn't say it wasn't worth trying. I just said that it didn't work too well for me.
Your median number of members who share your books is 45, so I don't think your library is too obscure for similar libraries to be useful. It pretty reliably finds me libraries of interest, and my median is 30.
I hadn't seen that statistic. Interesting. I'll have to find that statistic. :-)
But I've looked at my similar libraries. The first one on my list is moderately interesting (and in fact is someone whom I have met, decades ago), but the next few are not.
The real reason for my statement is that I looked at the library most like mine (https://www.librarything.com/profile/dankeding). His library is the closest to mine -- but, as of the last time I looked, I was only the seventeenth-closest to his. It's long branch assimilation -- my library is so weird that a library only moderately like mine comes up closest.
Possibly the high median score is because I have a lot of books that a lot of others have (many editions of The Iliad, The Odyssey, The Complete Works of Shakespeare, sundry Bible editions, a few popular fantasy works and common references). But these are not the interesting parts of my library. The interesting parts are things like Traditional singers and songs from Ontario (three other copies), The bothy songs and ballads of Aberdeen, Banff and Moray, Angus and the Mearns (three other copies), Folksongs of Mississippi and their background (one other copy).
Still, I didn't say it wasn't worth trying. I just said that it didn't work too well for me.
36spiphany
>35 waltzmn: Possibly you could use the collections feature to better narrow in on libraries which have the same sorts of obscure titles that you do; or rather, to filter out popular titles that may be skewing the results -- i.e., put Homer, Shakespeare, the Bible and whatnot in a separate collection and uncheck the option to use that collection for the purposes of connections.
37waltzmn
>36 spiphany:
Possibly you could use the collections feature to better narrow in on libraries which have the same sorts of obscure titles that you do; or rather, to filter out popular titles that may be skewing the results -- i.e., put Homer, Shakespeare, the Bible and whatnot in a separate collection and uncheck the option to use that collection for the purposes of connections.
Interesting thought. I never considered skewing my collections that way. :-) Thanks.
Possibly you could use the collections feature to better narrow in on libraries which have the same sorts of obscure titles that you do; or rather, to filter out popular titles that may be skewing the results -- i.e., put Homer, Shakespeare, the Bible and whatnot in a separate collection and uncheck the option to use that collection for the purposes of connections.
Interesting thought. I never considered skewing my collections that way. :-) Thanks.
38AndreasJ
>35 waltzmn:
I guess we just want different things from “similar libraries”. I don’t even show up under the (weighted) Members With Your Books for the library closest to mine, but I don’t see that as a problem.
The statistic is under Charts & Graphs > Books > Cataloging, BTW.
I guess we just want different things from “similar libraries”. I don’t even show up under the (weighted) Members With Your Books for the library closest to mine, but I don’t see that as a problem.
The statistic is under Charts & Graphs > Books > Cataloging, BTW.
39sarahemmm
I've just read through this thread. Maybe I don't look at books with very large numbers of reviews, because I find that I can very quickly skip down the reviews (which I usually sort by thumbs up), going past any one-liners and ones which just recount the plot.
I do take care to thumbs up good reviews, even if I don't care for the book, as I appreciate the work that has gone into them.
I do take care to thumbs up good reviews, even if I don't care for the book, as I appreciate the work that has gone into them.
40elenchus
Enjoying the discussion here.
I'll add that another issue for me, separate for the "weird library" issue that @waltzmn has raised, is for works that have scores if not hundreds of reviews. For a work I'm looking to learn something about, it would be helpful to filter all those reviews (even those with thumbs) based on criteria relevant to me.
This is another use case that would benefit from many of the same functionality already mentioned.
I'll add that another issue for me, separate for the "weird library" issue that @waltzmn has raised, is for works that have scores if not hundreds of reviews. For a work I'm looking to learn something about, it would be helpful to filter all those reviews (even those with thumbs) based on criteria relevant to me.
This is another use case that would benefit from many of the same functionality already mentioned.
41waltzmn
>40 elenchus:
I'll add that another issue for me, separate for the "weird library" issue that waltzmn has raised, is for works that have scores if not hundreds of reviews. For a work I'm looking to learn something about, it would be helpful to filter all those reviews (even those with thumbs) based on criteria relevant to me.
This is another use case that would benefit from many of the same functionality already mentioned.
That's actually the place I started from, though I don't remember if it was here or the Early Reviewers thread. I was trying to figure out every way I could to get closer to that goal: Both filters based on precise criteria such as length and somewhat fuzzier ones such as similarity to one's own interests.
But a "similarity rating" could be useful even for books with not many reviews: One might read the sole review in a different light if you knew that the reviewer generally agreed with you or not.
---
BTW, I meant to post this next part earlier, but it didn't show up; I must have pressed the wrong button (it's been that sort of day :-).
>38 AndreasJ:
I guess we just want different things from “similar libraries”. I don’t even show up under the (weighted) Members With Your Books for the library closest to mine, but I don’t see that as a problem.
To be clear, I'm just fine with having a funny library. I like being weird. :-) All I'm saying is that, if one doesn't have any particularly similar libraries, using it as a statistical basis won't be as effective.
Thanks for the tip about the median number of people with one's books.
I'll add that another issue for me, separate for the "weird library" issue that waltzmn has raised, is for works that have scores if not hundreds of reviews. For a work I'm looking to learn something about, it would be helpful to filter all those reviews (even those with thumbs) based on criteria relevant to me.
This is another use case that would benefit from many of the same functionality already mentioned.
That's actually the place I started from, though I don't remember if it was here or the Early Reviewers thread. I was trying to figure out every way I could to get closer to that goal: Both filters based on precise criteria such as length and somewhat fuzzier ones such as similarity to one's own interests.
But a "similarity rating" could be useful even for books with not many reviews: One might read the sole review in a different light if you knew that the reviewer generally agreed with you or not.
---
BTW, I meant to post this next part earlier, but it didn't show up; I must have pressed the wrong button (it's been that sort of day :-).
>38 AndreasJ:
I guess we just want different things from “similar libraries”. I don’t even show up under the (weighted) Members With Your Books for the library closest to mine, but I don’t see that as a problem.
To be clear, I'm just fine with having a funny library. I like being weird. :-) All I'm saying is that, if one doesn't have any particularly similar libraries, using it as a statistical basis won't be as effective.
Thanks for the tip about the median number of people with one's books.
42reading_fox
Just to add - and I've only skimmed so apologies if it's been raised before - that the vast majority of books have no/1 reviews. I think the default view if 4 reviews, and it's easy to skim that to see if any of them are relevant to you. I don't think a filter is actually going to be that useful most of the time, but tt would be nice to highlight in some way that a review is by a library you've marked as interesting.
(it is of course already possible to see the reviews that a specific library has written, but this is of less use if you're looking for a review of a specific book).
(it is of course already possible to see the reviews that a specific library has written, but this is of less use if you're looking for a review of a specific book).
43waltzmn
>42 reading_fox:
Just to add - and I've only skimmed so apologies if it's been raised before - that the vast majority of books have no/1 reviews. I think the default view if 4 reviews, and it's easy to skim that to see if any of them are relevant to you. I don't think a filter is actually going to be that useful most of the time, but tt would be nice to highlight in some way that a review is by a library you've marked as interesting.
This raises another thought. For the most part, you are right; most of my books have 0 or 1 reviews. For what it's worth, on my first page, the numbers are 1-1-0-20-0-0-0-0-0-1-0-2-0-0-0-1-0-1-0-0.
And the very first one isn't a review; it's another of these location indications. :-(
But it would have been useful to sort the one with 20 based on length. A lot of them are pretty feeble.
However, the real issue is on the next page. There, I find a book with 302 reviews. Which isn't too surprising, since the book is Beowulf.
The reason I bring this up is because I have fourteen editions of Beowulf (one Old English without any Modern English version, one Old English with glosses, one Old English with a modern parallel, the rest translations. Some are part of anthologies and hardly count).
Am I interested in more editions? Possibly, but I would need to be convinced. And most of those reviews are of different translations than I am interested in. If I am looking for a review, I am looking for a review of the Seamus Heaney translation, or the J. R. R. Tolkien translation, or the Howell D. Chickering Old English version, or whatever, not of Beowulf as a book. Odds are that I already know more about it than at least 290 of those 302 reviewers. :-)
Filters to let one search by translation or edition would really, really help. Lots of very popular books with multiple translations where you want to search through the translations. Just start with the Iliad and Odyssey and Aeneid and keep going from there.
Just to add - and I've only skimmed so apologies if it's been raised before - that the vast majority of books have no/1 reviews. I think the default view if 4 reviews, and it's easy to skim that to see if any of them are relevant to you. I don't think a filter is actually going to be that useful most of the time, but tt would be nice to highlight in some way that a review is by a library you've marked as interesting.
This raises another thought. For the most part, you are right; most of my books have 0 or 1 reviews. For what it's worth, on my first page, the numbers are 1-1-0-20-0-0-0-0-0-1-0-2-0-0-0-1-0-1-0-0.
And the very first one isn't a review; it's another of these location indications. :-(
But it would have been useful to sort the one with 20 based on length. A lot of them are pretty feeble.
However, the real issue is on the next page. There, I find a book with 302 reviews. Which isn't too surprising, since the book is Beowulf.
The reason I bring this up is because I have fourteen editions of Beowulf (one Old English without any Modern English version, one Old English with glosses, one Old English with a modern parallel, the rest translations. Some are part of anthologies and hardly count).
Am I interested in more editions? Possibly, but I would need to be convinced. And most of those reviews are of different translations than I am interested in. If I am looking for a review, I am looking for a review of the Seamus Heaney translation, or the J. R. R. Tolkien translation, or the Howell D. Chickering Old English version, or whatever, not of Beowulf as a book. Odds are that I already know more about it than at least 290 of those 302 reviewers. :-)
Filters to let one search by translation or edition would really, really help. Lots of very popular books with multiple translations where you want to search through the translations. Just start with the Iliad and Odyssey and Aeneid and keep going from there.
44AnnieMod
>43 waltzmn: The problem with this is that LT does not have an editions/translations layer - you essentially do not know what version someone has unless they add some identifying information OR tell you so in the review. (don't point out to the editions page on the work level - these are technical editions (aka non-splittables) and not what you are asking for - a single ISBN can have a lot of lines on that editions page). The best LT can do is to do a filter based on ISBN I guess (and that will work if people use the correct ISBN (which do not always happen), you know all the ISBNs of the editions you are looking for and the books have ISBNs to start with). You will struggle even to get only reviews in English if a member of the site had not set their review language right (or if they review in a language different from the language of the book they read - I do this often when not reading in English for example).
45norabelle414
It would be nice if something happened to blue flag reviews besides just dropping to the bottom when sorted by votes. Like if they did not count toward the total number of reviews for that work, or didn't show up on the Zeitgeist "first reviews" page.
46waltzmn
>44 AnnieMod:
The problem with this is that LT does not have an editions/translations layer
I know. That's my point. No way to search for, say, Burton Raffel's edition of Beowulf. But if one could look at only reviews of Beowulf that contain the word "Raffel," or "Heaney," or whatever, you'll get mostly relevant reviews.
It won't be perfect -- having the various translations and editions separated would be the ideal answer -- but having the filter would help. For a book with 302 reviews, it's a very important function.
The problem with this is that LT does not have an editions/translations layer
I know. That's my point. No way to search for, say, Burton Raffel's edition of Beowulf. But if one could look at only reviews of Beowulf that contain the word "Raffel," or "Heaney," or whatever, you'll get mostly relevant reviews.
It won't be perfect -- having the various translations and editions separated would be the ideal answer -- but having the filter would help. For a book with 302 reviews, it's a very important function.
47AnnieMod
>46 waltzmn: But they may not contain the word - if I had filled my secondary authors, I may not mention which translation I am writing about because it is my review and I have it on my page :) Or not call it with the name you expect.
I get what you are saying but it won't help much (outside of a few works - and even there... see above). Developing a feature that caters to a very small percentage of the works is not something I can see happening very soon. Just saying. Plus... "View All Reviews" and using the browser search functionality will get you the same even now - I do that when I really look for editions reviews. Unless you want the filter to look at secondary authors and then you are relying on people adding them and it gets even less likely to happen.
I get what you are saying but it won't help much (outside of a few works - and even there... see above). Developing a feature that caters to a very small percentage of the works is not something I can see happening very soon. Just saying. Plus... "View All Reviews" and using the browser search functionality will get you the same even now - I do that when I really look for editions reviews. Unless you want the filter to look at secondary authors and then you are relying on people adding them and it gets even less likely to happen.
48waltzmn
>47 AnnieMod:
Developing a feature that caters to a very small percentage of the works is not something I can see happening very soon.
I think you dismissing this too easily. There is potential for much more. Yes, that is a particular type of search that won't be done very often. But there are many use cases for a strong search function for reviews. Let me give two more examples.
The first is again from Beowulf. I probably have enough translations of Beowulf. But only a few of them have information about the sole, damaged, manuscript of the text: British Library Cotton Vitellius A.xv. So to learn if a review talks about the thing I really want, I might search reviews of Beowulf for references to "Cotton Vitellius."
Now your show all reviews/search trick would work for that -- if the review uses those exact words in that exact order. But I don't know if the review will mention "Cotton Vitellius" or "Cotton Library" or "Cotton manuscript" or "Vitellius manuscript," and the words "Cotton" and "Vitellius" may not be consecutive (the manuscript might be called "Cotton MS Vitellius" or the like. And an old book would refer to the "British Museum," not the "British Library"). So what I want to do is search Beowulf reviews for "Cotton" AND "Vitellius" OR "Library" OR "Manuscript." Can't do that with a browser.
Similarly, one of the two oldest manuscripts of a substantial reference to Robin Hood ("Robin Hood and the Potter") is in Cambridge, Cambridge University Library MS. Ee.4.35. This manuscript contains a merchant's mark and an owner's name, Richard Calle. (Well, Ricardo Calle, since it's in Latin, but his name in English was Richard Calle.)
That is a very important manuscript, so information about the owner is significant. But who is Richard Calle? The manuscript doesn't tell us, and since it's from the fifteenth century, there are no census records to check. But there is one other source: the famous Paston Letters. Richard Calle was for many years the Pastons' chief business agent; he married one of their daughters (to the fury of the other Pastons, who opposed the match, but that's beside the point). So how can one find out about Calle? By reading books about the Pastons. So, again, I want to do a complex search of reviews, looking for "Richard" or "Ricardo" and "Calle" or "Call" (the other spelling of his name).
Ideally, I'd even like to do a search like that on all reviews, to find out about books on Richard Calle that I didn't know about. (No, I can't search the site for that; searching for "Richard Calle" got me 953 hits, and at least the first few irrelevant; searching for "Richard Calle Pastons" got no hits at all. Searching my own library got me 11 hits, 10 of them false positives.)
(That, incidentally, is a third feature request: The ability to search all reviews.)
So we have many identified uses for a review search mechanism, not just one:
1. To look for particular translations of a work translated repeatedly
2. To look for particular aspects of a review, such as references to a particular person or manuscript
3. To look for reviews by (or NOT by) a particular person/interesting library
4. To look for reviews by those with similar libraries
5. To look for those with a certain minimum length
And I may have forgotten some that others suggested.
Individually, probably, all of these are of minor use. But collectively they are more significant. And if you build the feature, someone may well find other ways to use it.
Developing a feature that caters to a very small percentage of the works is not something I can see happening very soon.
I think you dismissing this too easily. There is potential for much more. Yes, that is a particular type of search that won't be done very often. But there are many use cases for a strong search function for reviews. Let me give two more examples.
The first is again from Beowulf. I probably have enough translations of Beowulf. But only a few of them have information about the sole, damaged, manuscript of the text: British Library Cotton Vitellius A.xv. So to learn if a review talks about the thing I really want, I might search reviews of Beowulf for references to "Cotton Vitellius."
Now your show all reviews/search trick would work for that -- if the review uses those exact words in that exact order. But I don't know if the review will mention "Cotton Vitellius" or "Cotton Library" or "Cotton manuscript" or "Vitellius manuscript," and the words "Cotton" and "Vitellius" may not be consecutive (the manuscript might be called "Cotton MS Vitellius" or the like. And an old book would refer to the "British Museum," not the "British Library"). So what I want to do is search Beowulf reviews for "Cotton" AND "Vitellius" OR "Library" OR "Manuscript." Can't do that with a browser.
Similarly, one of the two oldest manuscripts of a substantial reference to Robin Hood ("Robin Hood and the Potter") is in Cambridge, Cambridge University Library MS. Ee.4.35. This manuscript contains a merchant's mark and an owner's name, Richard Calle. (Well, Ricardo Calle, since it's in Latin, but his name in English was Richard Calle.)
That is a very important manuscript, so information about the owner is significant. But who is Richard Calle? The manuscript doesn't tell us, and since it's from the fifteenth century, there are no census records to check. But there is one other source: the famous Paston Letters. Richard Calle was for many years the Pastons' chief business agent; he married one of their daughters (to the fury of the other Pastons, who opposed the match, but that's beside the point). So how can one find out about Calle? By reading books about the Pastons. So, again, I want to do a complex search of reviews, looking for "Richard" or "Ricardo" and "Calle" or "Call" (the other spelling of his name).
Ideally, I'd even like to do a search like that on all reviews, to find out about books on Richard Calle that I didn't know about. (No, I can't search the site for that; searching for "Richard Calle" got me 953 hits, and at least the first few irrelevant; searching for "Richard Calle Pastons" got no hits at all. Searching my own library got me 11 hits, 10 of them false positives.)
(That, incidentally, is a third feature request: The ability to search all reviews.)
So we have many identified uses for a review search mechanism, not just one:
1. To look for particular translations of a work translated repeatedly
2. To look for particular aspects of a review, such as references to a particular person or manuscript
3. To look for reviews by (or NOT by) a particular person/interesting library
4. To look for reviews by those with similar libraries
5. To look for those with a certain minimum length
And I may have forgotten some that others suggested.
Individually, probably, all of these are of minor use. But collectively they are more significant. And if you build the feature, someone may well find other ways to use it.
