Lois’s answer to “Not really a question, but something I thought you might find interesting. Based on my buying Priso…” > Likes and Comments
8 likes · Like
That's probably a misread of Vallista by Brust. Just FYI.
Well, since I'm a huge LMB fan (Paladin of Souls is one of my favorite books EVER), hearing about that is a personal thrill for me! :)
I am guessing "interiority" means self awareness or perhaps internal monologue?
I found a page that has various possible ways Amazon runs their purchase suggestions. There's talk of patents and market analysis, and 'pipelines', algorithms and neural networks. Which, some of it sounds a whole lot like "I don't know". But, there it is. ^^
This is the link: https://stackoverflow.com/questions/2...
... The list sounds like something concocted hastily from some very broad genre category ("women friendly fantasy" !?!?!?!?!??!) and currently "trending" titles which need to be promoted.
As far as I know, their main source for recommendations is "some people bought these books together". And it seems to work pretty well.
At a guess, I suspect their algorithms are similar to the word-cloud generation ones. Mostly one picks terms (such as woman and fantasy and many others), matches blurbs or other information about a book to score how much "woman-ness" and "fantasy-ness" is in their corpus of books, then plays clustering algorithm games to decide which ones are similar to each other.
The results of such techniques can do surprisingly well, and can do surprisingly badly. In this case, one factor on the bad side is that blurbs tend to misrepresent books half the time anyway.
I'm sure it is a relatively simple algorithm that looks at purchase and viewing commonalities - so if people who buy your books commonly buy those others, or at least read through them, they get grouped together. I imagine purchases with high reviews count more strongly, and so on.
FYI - Steven Brust's book is Vallista, not Ballista.
The very similar list of recommendations for me on the Limnos page says "other people who bought this book also bought...".
I don't know the algorithm, but you have to have at least 25 reviews before your books can be considered.
I think it’s the same general idea as this map—what people who bought that book also bought or maybe liked in reviews. But the map is more fun. http://www.literature-map.com/lois+mc...
Pity it's a mistook. I would like to read Penric's Ballista
back to top
date
newest »
newest »
message 1:
by
Jacob
(new)
Nov 11, 2017 08:19PM
That's probably a misread of Vallista by Brust. Just FYI.
reply
|
flag
Well, since I'm a huge LMB fan (Paladin of Souls is one of my favorite books EVER), hearing about that is a personal thrill for me! :)
I am guessing "interiority" means self awareness or perhaps internal monologue?
I found a page that has various possible ways Amazon runs their purchase suggestions. There's talk of patents and market analysis, and 'pipelines', algorithms and neural networks. Which, some of it sounds a whole lot like "I don't know". But, there it is. ^^This is the link: https://stackoverflow.com/questions/2...
... The list sounds like something concocted hastily from some very broad genre category ("women friendly fantasy" !?!?!?!?!??!) and currently "trending" titles which need to be promoted.
As far as I know, their main source for recommendations is "some people bought these books together". And it seems to work pretty well.
At a guess, I suspect their algorithms are similar to the word-cloud generation ones. Mostly one picks terms (such as woman and fantasy and many others), matches blurbs or other information about a book to score how much "woman-ness" and "fantasy-ness" is in their corpus of books, then plays clustering algorithm games to decide which ones are similar to each other.The results of such techniques can do surprisingly well, and can do surprisingly badly. In this case, one factor on the bad side is that blurbs tend to misrepresent books half the time anyway.
I'm sure it is a relatively simple algorithm that looks at purchase and viewing commonalities - so if people who buy your books commonly buy those others, or at least read through them, they get grouped together. I imagine purchases with high reviews count more strongly, and so on. FYI - Steven Brust's book is Vallista, not Ballista.
The very similar list of recommendations for me on the Limnos page says "other people who bought this book also bought...".
I don't know the algorithm, but you have to have at least 25 reviews before your books can be considered.
I think it’s the same general idea as this map—what people who bought that book also bought or maybe liked in reviews. But the map is more fun. http://www.literature-map.com/lois+mc...
Pity it's a mistook. I would like to read Penric's Ballista
