Slashdot Mirror

← Back to Stories (view on slashdot.org)

New Algorithm Recognizes Both Good and Bad Fake Reviews (thestack.com)

Posted by samzenpus on from the do-you-really-like-it? dept.

An anonymous reader writes: Researchers from the university of Sao Paolo have developed an algorithm able to identify both good and bad online reviews in the massive daily chatter of millions of peer-community posts, and in lateral mendacities at social network sites such as Google+ and Facebook reposts and 'likes'. Two of the datasets tested in the research were from Amazon, which has a vested interest in restoring the reputation of its community reviews, and has recently taken action on the matter.

27 of 59 comments (clear)

  1. One modest opinion by Marginal+Coward · · Score: 1

    With Siskel and Ebert now gone, my bot is picking up the slack: it gives this concept two thumbs down.

  2. Great algorithm - would implement! by Anonymous Coward · · Score: 5, Funny

    While I have spent a lot of time studying algorithms, I still like to look for new algorithms from time to time. This algorithm is definitely one of the better algorithms I have come across in my life.

    When I first started studying the algorithm, I wasn't sure whether it was for me. However, after the first few pages, the algorithm began to grab my attention, and soon I was finding its average and worst case complexity and formally proving the correctness of several subroutines. I was astounded to find that it was at least as good as all the other algorithms I'd found over my twenty years' experience.

    While I wouldn't recommend this algorithm to a beginner without guidance, the journeyman will soon find themselves reaching master by applying this algorithm. The master cannot fail to learn something new from it.

    1. Re:Great algorithm - would implement! by Hognoxious · · Score: 1

      +1 Would buy a new keyboard again.

      --
      Confucius say, "Find worm in apple - bad. Find half a worm - worse."
    2. Re:Great algorithm - would implement! by Marginal+Coward · · Score: 1

      Hey, that's pretty good! But it seems just a bit too...well...polished if you're actually satirizing something like this:

      hey mama bebs, first of all thanks for drppion on my super unattended blogs and I may not be the perfect person to send replies back the soonest its better late than not doing it all. Anyhoo..we have our disagreements in life,perspectives etc etc..but nevertheless when it concerns to friendship it is always seemed hard to explain. Whether we like it or not, the truth hurts. But just do what you love to do and don’t expect things turn out that way it supposed to be is a great thing to do..let’s talk. have a great one!

      The preceding is actual comment spam from one of my websites. Evidently, the bot was programmed by someone who doesn't speak English very goodly.

    3. Re:Great algorithm - would implement! by TiggerTheCat · · Score: 2

      I wonder if this, or a variant, could pick out forum spys? http://pastebin.com/irj4Fyd5

    4. Re:Great algorithm - would implement! by TiggerTheCat · · Score: 1

      Actually, I wonder if some forum spies are dissing this algorithm so it won't be used to find forum spies?

    5. Re:Great algorithm - would implement! by Hognoxious · · Score: 1

      I think I've seen that before. It was in a forum about C++.

      --
      Confucius say, "Find worm in apple - bad. Find half a worm - worse."
    6. Re:Great algorithm - would implement! by labnet · · Score: 1

      I bought this algorithm based on Anonymous Cowards review; and while I was at first disjointed I persisted until it gave results like no other algorithm before it. While parts of me may still be unhappy, the rest of me is feeling very fulfilled that I made the right algorithm choice. Thanks Anonymous.

      --
      46137
    7. Re:Great algorithm - would implement! by BoogieChile · · Score: 1

      Algorithm contained bobcat. Would not implement again, although I could only be impressed by how they managed to get it to fold up so small.

  3. Poor wording... by SofiKadaj · · Score: 1

    New Algorithm Recognizes Both Good and Bad (Fake Reviews): The algorithm recognizes good fake reviews and bad fake reviews, but not real reviews.
    New Algorithm Recognizes Both Good and (Bad Fake) Reviews: The algorithm recognizes good reviews and bad fake reviews, but no real bad reviews or fake good reviews.
    New Algorithm Recognizes Both (Good and (Bad Fake) Reviews): The algorithm recognizes a particular good review and a particular bad fake review.
    New Algorithm Recognizes Both (Good and Bad (Fake Reviews)): The algorithm recognizes a particular good review and a particular bad review.
    New Algorithm Recognizes Both Good and Bad Fake Reviews: The algorithm recognizes ruthlessly pedantic slashdot comments.

    1. Re:Poor wording... by SeaFox · · Score: 1

      Would you like a review?

  4. That's easy by Anonymous Coward · · Score: 2, Funny

    Just return true all the time. Sure there will be a few false positives but not enough to throw the results off by much.

  5. My Review of this article by Anonymous Coward · · Score: 2, Funny

    It combines orgasmic whitefish essences and loud mustard overtones with a stale nacho aftertaste.

    1. Re:My Review of this article by Austerity+Empowers · · Score: 1

      I plan to plagiarize this for an audiophile review, just FYI.

  6. What about obvious fake humorous reviews by hackwrench · · Score: 1

    Part of what makes me come back to Amazon's site to buy products is the humorous fake reviews for things that are absurdly expensive or seems like it shouldn't exist.

    1. Re:What about obvious fake humorous reviews by rtb61 · · Score: 3, Insightful

      Don't worry Amazon is just interest in rebuilding the 'reputation' of their reviews and not in the actual real genuiness or accuracy of the reviews, they just want to believe the reviews and buy junk. So they are basically engaging in a public relations exercise to give their reviews a positive review because so many people have given up on IMDB because the publicly accepted default review is a 10 out of 10 by a one off 'er' reviewer, so bad that it IMDB unusable as a movie review system. Basically skip the first bunch of pages (seriously up to and over 10 pages) as PR bullshit to get to real actual reviews.

      --
      Chaos - everything, everywhere, everywhen
  7. Re:Moderation by CCarrot · · Score: 4, Interesting

    Why can't we just have moderation? Most of the poor (low quality, as opposed to negative) reviews are easy to spot; knuckleheads that break things two seconds after they get it out of the package; half the things they buy are mysteriously "DOA." Moderation enables your smart customers to punish your idiot customers.

    Just spotted one of these today on eham.net. A beautiful hand made band pass filter rated 4.2/5 instead of the 5/5 it deserves because — years ago — some moron used it on the wrong band for the wrong purpose and rated it 0/5.

    I thought that was what the "Was this review helpful to you? YES/NO/Report Abuse" options were for?

    Effective community policing needs an involved community like slashdot. Even here, where nobody's trying to sell you anything other than an opinion, the sock-puppet modding can get out of hand sometimes...

    --
    "I love animals! Some are cute, others are tasty, what's not to like?" - Betsy Schroeder, Jeopardy contestant
  8. Fake Reviews on Amazon by ameoba-light · · Score: 1

    If you haven't seen this, there is a service that looks at reviews to find if they are fake or not: http://fakespot.com/ From what I can gather about them, they also use clustering algorithms to detect the fake Amazon reviews.

    1. Re:Fake Reviews on Amazon by BasilBrush · · Score: 1

      Useless. I directed it to a book with 467 reviews. It claimed where were not enough reviews to analyse.

  9. The name is by Anonymous Coward · · Score: 1

    actually "Universidade de São Paulo", located in the city and state of the same name. I get that you anglophones don't go well with tildes [and for that you are excused], but don't change so drastically the name of my town/state.

  10. "restoring the reputation" by taustin · · Score: 1

    The only way Amazon could restore the reputation of their reviews would be to change from the bad it's always been to something better (because it couldn't get any worse), then make it bad again.

  11. This is amazing! by LordHighExecutioner · · Score: 1

    *** this post has been recognized as fake and removed by the new automatic fake post identifier ***

  12. Re:Um. by KGIII · · Score: 1

    Well, you're obviously trolling, at least a little, but still deserve a reply. Any deterministic algorithm can be bested simply by determining what the algorithm does and then doing something different. So, yes, this isn't actually going to work as a permanent solution. It may work on past reviews, it may work for some time in the future, and it may work against those unwilling to put an effort into their work. However, someone will, almost certainly, and find the flaws. There is a technical term for this but, suffice to say, it boils down to there being no such thing as a free lunch.

    Maths, explained conceptually, are a good thing. *nods* I'm not good at analogies but think of it like a law. No law can be without loopholes or it would be draconian and not have a desirable outcome. Yet, each loophole have potential abuses. There is no such thing as a perfect law which will give you ideal results in all situations.

    Or, if you'd prefer, an automobile that is optimized for efficiency will not be good for hauling around extra weight - it will even be less efficient than an automobile designed to carry such a load. One can not create an automobile that will function ideally in all situations. Even if you could create such a thing, someone would just come along with a different use-case where such failed. It may take someone a moment to find that use-case but it will happen.

    --
    "So long and thanks for all the fish."
  13. Obligatory XKCD by RoloDMonkey · · Score: 1

    https://xkcd.com/810/

    --
    Long live the Speaker Bracelet
    Rolo D. Monkey
  14. Re:Um. by BasilBrush · · Score: 1

    And yet spam email detection has got so good that I almost never get a spam.

    And my ad-blocker is so good that I almost never see an ad.

    So I don't think much to your theory of inevitable defeat.

  15. Re:Um. by KGIII · · Score: 1

    No? Then why'd you use the word "almost?" Did you block APK's ad yet? Does your spam filtering never get a false positive, never let spam through?

    And no, it's not my theory. You could see it as an off-shot of things like this: https://en.wikipedia.org/wiki/...

    It's a pretty basic idea and doesn't really require that you agree with it. If you can prove them wrong then, well, you'd be famous.

    --
    "So long and thanks for all the fish."
  16. Re:Um. by BasilBrush · · Score: 1

    Because almost is very useful. If a site like Amazon can kill 99% of fake reviews that's an extremely worthwhile thing to do. It doesn't have to be perfect.

    And spam filtering does a hell of a lot better than 99% these days. I honestly can't remember when I last got a real email spam. It may be years.

    And no, it's not my theory. You could see it as an off-shot of things like this: https://en.wikipedia.org/wiki/..

    Then either it's a crap theory or you are misusing it. But there's no fame in pointing out that spam filtering works. It's surprising you haven't realised it.