Kindle Forum banner

1 - 11 of 11 Posts

·
Registered
Joined
·
354 Posts
Discussion Starter #1

·
Registered
Joined
·
2,482 Posts
Fakespot has been discussed here and the consensus is that it yields far too many false positives, and is not reliable.

Anyone know anything out the other tools?
 

·
Registered
Joined
·
1,454 Posts
David VanDyke said:
Anyone know anything out the other tools?
No inside knowledge, but I gave both a quick try on a nonfic writing book and a novel whose authors I consider beyond reproach and with reviews that look typical.

Both handled the novel fine.

Reviewmeta passed the nonfic with a few concerns. Fakespot graded it "F" and said the true rank is likely 1/2 star rather than 4.5. This looks like an extreme false positive.

And in presentation, I prefer the Reviewmeta methodology where it points out individual factors for why it gives the warnings or failures. This granularity makes it more valuable since you can sanity-check areas of concern yourself. Seems like a decent tool.
 

·
Super Moderator
Joined
·
10,088 Posts
David VanDyke said:
Fakespot has been discussed here and the consensus is that it yields far too many false positives, and is not reliable.
Yup. The most recent thread was about a month ago, I think.
 

·
Registered
Joined
·
1,454 Posts
Shelley K said:
Fakespot is hilarious. I just checked one of my books with over five dozen organic reviews. It says 35% of them are fake, and gives it a D for unreliable reviews. Part of the reason for that is the overwhelmingly positive bent. So if people like your book, Fakespot thinks you're cheating. I don't need to point out how stupid that is.
It's a really poor assumption for books with ~100 reviews or less. It's likely based on a statistical average scraped from a broad cross section of products that makes sense in the larger context. ( <-- benefit of the doubt mode)
 

·
Registered
Joined
·
354 Posts
May have some degree of accuracy for other products, but is so stupid regarding books that it should be illegal for them to do it. VERY misleading information.
 

·
Registered
Joined
·
6,318 Posts
J. Tanner said:
It's a really poor assumption for books with ~100 reviews or less. It's likely based on a statistical average scraped from a broad cross section of products that makes sense in the larger context. ( <-- benefit of the doubt mode)
I got an F on a book with over 1,000 Amazon US reviews. My first book, published when I knew flat nobody and nothing and just put the book up there to see what happened. F. I tried a few other books with similar results. I don't think Fakespot's too reliable. :)
 

·
Registered
Joined
·
1,454 Posts
JRTomlin said:
ETA: I just tested ReviewMeta (which I had tested before) with The Templar's Cross for which I have NEVER solicited reviews. It has 312 reviews and they claimed that 100+ were fake because of 'unverified purchases' (KU ... duh) and repeated phrases such as 'like historical fiction', "15th century Scotland" and "medieval Scotland'. Again... duh. Various authors have given them feedback on how to improve their rating but they prefer to just stick a lame warning on it.
KU doesn't help for sure (surprise!), but I went ahead and checked your book and if you ignore the overall (silly) Fail rating and look at the granular details the book does fine. Their modifier shifts your book from 3.8 to 3.7 stars. So they don't think the unverified reviews are impacting the results much. They really should have categorized that as a Pass.

Their system lets you really investigate the "why" behind the answer it spits out, and I consider that a pro despite the fact that it's not handling books as well as it might.
 

·
Registered
Joined
·
2,767 Posts
I know it when I see it. I don't need bad algorithms to work that out for me. It is easy enough to pinpoint the books which were marketed with ARC and street teams, and it is also easy to see which have shady aka solicited/bought reviews. Both show me I shouldn't be buying that book regardless of how much hype it gets.
 

·
Registered
Joined
·
4,299 Posts
I find it interesting that both analysis tools find a problem with "most repeated word" in my reviews. That word, of course, is "book".
 
1 - 11 of 11 Posts
Top