Recommendations on Go Set a Watchman?

ConversesTalk about LibraryThing

Afegeix-te a LibraryThing per participar.

Recommendations on Go Set a Watchman?

Aquest tema està marcat com "inactiu": L'últim missatge és de fa més de 90 dies. Podeu revifar-lo enviant una resposta.

1timspalding
Editat: jul. 11, 2015, 12:06 pm

What would good recommendations on Go Set a Watchman look like?

Algorithms fail because it's basically "people who keep up with 'current' books."

Here's what we have now: http://www.librarything.com/work/15706335/recommendations

Or maybe that is the answer. I don't think most of its readers want to go deep on, say, 1950s fiction touching similar topics. Other "event" books? Books mostly about Harper Lee and To Kill a Mockingbird?

2lilithcat
jul. 11, 2015, 10:17 am

>1 timspalding:

That "athena" link requires authorization.

3Jarandel
Editat: jul. 12, 2015, 1:20 pm

I've seen that sort of trend for some of my own books as well, though it was more "people who buy/read current fantasy" or "people who buy/read current sci-fi", some books also definitely seem to have "people who buy/read recent prize winners", "people who buy/read recent prize winners and a more or less comprehensive amount of their short/extended lists", or "people who buy/read whatever the current buzz is about" patterns.

Not sure it really needs to be corrected, people reading with a preference for the bleeding edge of published books, or for social fodder is a thing, and hopefully as the buzz dies down the books that have any enduring appeal will increasingly be picked for their intrinsic traits, or trickle down better or worse among different tribes of readers, and recs shift accordingly.

4timspalding
jul. 11, 2015, 12:07 pm

>2 lilithcat:

Sorry. Fixed.

5Petroglyph
jul. 11, 2015, 3:15 pm

Considering that the book has not officially been published yet, I think the best (i.e. most honest) recommendations are indeed books that are similar in age, premise or topic. What else do you want to show up there?

6timspalding
jul. 11, 2015, 3:16 pm

Well, what are they?

7elenchus
jul. 11, 2015, 4:03 pm

I'd argue it's difficult to provide an answer with much integrity without having read it, and by definition few people have at this point. I wouldn't want LT to somehow "break the story" on "true" recommendations by guessing or whatever. And while a few LTers may have read it by now, that's just it: they are few.

I'm with Petroglyph on this one: wait until the data can be compiled. Why, apart from celebrity / headlines, are you looking to get a recommendation now that's any different from what LT is able to provide using the current algorithm?

8timspalding
jul. 11, 2015, 4:08 pm

>7 elenchus:

Yeah, but the data is going to be people who read hot books. We see this with other hot, current fiction. I mean, Gone Girl ( https://www.librarything.com/work/11234211/recommendations ) is no better.

9lilithcat
jul. 11, 2015, 4:40 pm

>7 elenchus:

Considering the extensive information available in published reviews and discussion of the book (see, for instance, today's New York Times), I'd say rational recommendations are certainly possible.

10jjwilson61
jul. 11, 2015, 5:20 pm

Are the tag-based recommendations better? Can you weight them more in the main recommendation algorithm?

11prosfilaes
jul. 11, 2015, 6:01 pm

>9 lilithcat: That's where member recommendations come in. It seems a little pointless to try and bend computer recommendations to try and do this job.

12MaureenRoy
Editat: jul. 11, 2015, 7:22 pm

July 11, 2015: The Guardian newspaper has an exclusive excerpt as of yesterday -- the complete first chapter of Go Set A Watchman:

http://www.theguardian.com/books/ng-interactive/2015/jul/10/go-set-a-watchman-re...

The Guardian also features actress Reese Witherspoon reading the first chapter aloud in an audio clip:

http://www.theguardian.com/books/2015/jul/10/listen-to-reese-witherspoon-reading...

Readers are then asked to comment on The Guardian website with their reactions to the book so far.

13Petroglyph
jul. 11, 2015, 7:55 pm

>8 timspalding:
Yes, people who keep up with recent books are the current input for whatever algorithms generate the recommendations, and the results will likely be of use to precisely those people. I don't see how this means that the algorithms have "failed", as you put it in >1 timspalding:. Once a more diverse group of members actually starts cataloguing the book, won't the recommendations shift accordingly?

Are you expecting the recommendations for this book to not change over time? Would you want the current recommendations to reflect the situation two months from now? What do you expect to see in the recommendations that is not showing up there now? I honestly don't understand the problem here.

>9 lilithcat:
What do you mean with "rational recommendations"? What is irrational about the current ones?

14lorax
jul. 11, 2015, 8:41 pm

Look, you're solving for an extreme corner case here. No algorithm is going to do a good job. Any tweaks you make to solve for this case are going to hurt the algorithm for most books. (You guys have done actual testing on your recommendations, beyond spot-checking a few books, right? Right?)

You know I've done recommendation engines for a living. Don't worry about the once-in-a-decade publishing (if not longer) event when you're tweaking them. Seriously, don't.

15timspalding
jul. 11, 2015, 9:29 pm

>14 lorax:

This is an edge case. But popular fiction is not. It's the hardest one.

16lorax
jul. 11, 2015, 10:37 pm

Okay, it wasn't obvious to me that you were using Go Set a Watchman as a standing for "popular literary fiction". I can see that being tough.

17MaureenRoy
jul. 12, 2015, 11:08 am

As of Sunday July 12th, opinions on the GSAW chapter 1 are all over the place. In my opinion, Harper Lee is incapable of writing a bad book. Today's comments:

http://www.theguardian.com/books/2015/jul/10/first-chapter-of-go-set-a-watchman-...