A university research team has developed an extraordinary computer algorithm to spot fake TripAdvisor reviews.
The development follows news that some properties are bribing guests with food and drink discounts to write positive reviews on TripAdvisor.
Posting creative and brilliant reviews on hotel review sites to bolster the standing of seedy or dodgy properties has become a nice little earner for some travellers. A thriving black market has sprung up, with squalid hotels giving discounts to guests who write glowing reviews. Despite such practices being illegal in many countries, an army of people is apparently willing to knock up a dishonest review for as little as AUD10.
The jig may now be up, however. Britain’s Daily Mail reports that researchers from Cornell University in the US have developed a fiendishly cunning algorithm that they claim can spot a fake review 90% of the time. They are still deciding whether to make the algorithm public.
The Cornell team started its research into what is termed “deceptive opinion spam” by asking freelance writers to produce 400 fake reviews of Chicago hotels and post them on review websites.
The research team then mixed the false reviews with 400 real reviews of the same hotels – and challenged three judges to tell the difference. The judges failed.
The researchers dug deeper and found giveaway signs, which they have incorporated into their algorithm. A narrative account of a holiday indicates the review may be fake. Other factors indicating a fake or “shill” review include excessive use of superlatives and lack of detail and description.
Travel review sites do their best to weed out “shill” reviews. Shill is an odd word, which entered the language in the 1920s from uncertain origins. It means “a person who publicises or praises something or someone for reasons of self-interest, personal profit, or friendship or loyalty”. Shills are attracted, for obvious reasons, to review sites.
Even if you eliminate shills, some reviews differ so widely that guests, and many hotel managers, are left scratching their heads (and not necessarily because of bedbugs). Some guests describe hotels as charming, friendly and delightful hideaways while other guests describe the same properties as smelly, vermin-infested hovels to be shunned at all costs.
A Cornell professor who worked on the fake-review-detection algorithm project said humans were used to talking face to face over the past 60,000 years and found it difficult to detect deception online.
TripAdvisor has reportedly blacklisted about 30 properties around the world for suspicious reviews. It may now be able to do something more concrete to weed those reviews out.
(Source: eglobaltravelmedia.com.au August 2011) Written by : Peter Needham
No comments:
Post a Comment