happytrailstoyou

Some ways to spot fake reviews: http://abcnews.go.com/Technology/sof...ry?id=14172704

HTTY

annesherrod

That is interesting. I have wondered how trip advisor determines if a review is fake.

The software is not flawless but sure can weed out some fakes I would assumeas the article states.

Thanks for posting.

ansh

This is MUST needed technology...specially for (what i use, i trust and i like - ) fodor's, tripadvisor, amazon etc.

Keith

"They took another 400 actual reviews of 20 leading Chicago hotels from TripAdvisor, the top travel site."

How many of THOSE were fake reviews?

LarryBee

Keith, you are right on. Although it is hard to know the detailed methods used to conduct this study, it seems flawed from the get go unless they can be very sure those 400 reviews were real. Further, I suspect that generally people who post fake reviews are either connected with the hotel (for a postive review) or have an axe to grind with the hotel (for a negative review) but either way, they likley are more familiar with the hotel than you would get from just reading an internet page. I doubt the 400 people assigned to write the known "fake" reviews fall into that category, so how can they be reprentative of who normally writes a fake review? Finally, and perhaps most damning, the 400 assigned to write the fake review did so for only ONE hotel, and that was compared against reviews from 20 different hotels! I don't doubt that software can be useful to "help" detect fake reviews, but I doubt it is as useful as this study suggests....

montereybob

When they took 400 "actual" reviews from Tripadvisor, how did they know they were
Actual"?

happytrailstoyou

Who knows, I wasn't at the Association for Computational Linguistica meeting when the study was presented there by a Cornell University doctoral student and his professor. I'm sure all that was explained there.

In any event, a review that includes specific information (large, clean, quiet rooms) is more useful than one that talks in generalities (everything was wonderful).

HTTY

LarryBee

I don't need a Cornell PhD to tell me a review that talks in specific terms is more useful than a review that talks in generalities. In any event, that was not the point of the study. Perhaps all was explained at the conference, but all we have to base comments on was the information given. I've worked in the biological sciences for decades and have attended and presented at many scientific conferences, and have written and served as a peer reviewer for many publications in scientific journals. Just because a study gets presented or published doesn't mean the correct methods were used or appropriate conclusions drawn. I'm not immediately convinced just hearing the words "Cornell" "doctoral" or "Computational Linguistica" thrown at me.

LarryBee

Hmm, just couldn't let it go so a quick web search turns up the actual paper:

http://www.cs.cornell.edu/~myleott/op_spamACL2011.pdf

Where I was wrong (or more accurately, misinformed by the news article): The study did indeed use generated fake reviews from 20 DIFFERENT hotels to compare to the "real" reviews from 20 area hotels. That is good. Where I was right: the description of how they detirmined a TRIPADVISOR review was "real" is unsatisficatory; they basically just assumed most of what is on tripadvisor is real-read it for yourself. Also, as I thought the people hired to write the fake reviews had no presumed knowledge of the hotel they were reviewing and were given only 30 minutes to write the fake review-is it any wonder that as a group these fake reviews had more in common with each other than either reviews that are real or fake reviews written carefully and deliberately by someone who hopes to profit by attracting more business? I'm not saying the comclusion is junk, I'd just like to see the study repeated, correcting the flaws pointed out, and see if the results are the same.

happytrailstoyou

LarryBee,

Thanks for your analysis. It seems the study is full of holes--like the sheets in a flop house.

HTTY