review spam detection via temporal pattern discovery
DESCRIPTION
Review Spam Detection via Temporal Pattern Discovery. Sihong Xie, Guan Wang, Shuyang Lin, Philip S. Yu Department of Computer Science University of Illinois at Chicago. What’s review spams. Give some examples of spam reviews Also give brief descriptions. - PowerPoint PPT PresentationTRANSCRIPT
Review Spam Detection via Temporal Pattern Discovery
Sihong Xie, Guan Wang, Shuyang Lin, Philip S. YuDepartment of Computer Science
University of Illinois at Chicago
What’s review spams•Created on review websites in order to
create positive impressions for bad products/stores, and make profit out of misled customers
•They are harmful: lead to poor customer experience, ruin reputation of good stores
•Guidelines to spot fake reviews for human, but it is hard for machines ([1])
Give some examples of spam reviewsAlso give brief descriptions
Give some examples of spam reviewsAlso give brief descriptions
http://consumerist.com/2010/04/how-you-spot-fake-online-reviews.html
Human friendly clues of spams
• Language features [5]
1.All praises2.say nothing about the product3.Red flag words4.mention the name a lot
Too hard for Too hard for machines:machines:
Involving natural Involving natural language processinglanguage processing
Machine friendly clues of spams• similar reviews (texts and ratings) on
one product / product group in a short time [1,2]
Machine friendly clues of spams• Group of spammers:
• Wrote reviews together frequently on the same set of products / stores [3]
Reviewer 1 Reviewer 2 Reviewer 3
How to play the spamming game
• Duplicated reviews: two reviews are almost the same
More More sophisticated sophisticated
writingswritingsEasy: shingling
Use different Use different reviewer idsreviewer ids
detection is easy [3]
Use different Use different reviewer idsreviewer ids
detection based on statistics of reviewer
Player 1Detection systems
Player 2Spammers
• Group spamming: a group of reviewers frequently write reviews together.
• Other kinds: targeting on the same product, similar texts/ratings by one id.
Failed machine friendlyclues of spams
If these reviews were posted by the same id, it would have been easy to detect
Same id wrote multiple reviews, making it easy to be detected. Smart spammers would avoid this
Reviewer 1 Reviewer 2 Reviewer 3
Failed machine friendlyclues of spams
Spammers like singleton spam
•Strong motivations to have singleton spams:
1.Need to boost the rating in a short time
2.Need to avoid being caught
• Post reviews with high rating under different names in a short time
Singleton reviews
0 +
Each reviewer id contributes only one review for one store only
A physical person can
register many reviewer ids
Reviewer id
Store
Singleton
non-singleton
+ Spammer
Registration
0 Normal reviewer
Facts of Singleton reviews
• Constitute a large portion of all the reviews
• Over 90% of the reviews are singleton reviews in this paper; similar situations in another dataset [4]
• More influential, more harmful
The challengesTraditional cluesTraditional clues shortcomingsshortcomings
Review features (bag of words, ratings, brand names reference) [4]
Hard for human, not to mention machines
Reviewer features (rating behaviors) [1]
Poor if one wrote only one review
Product/Store features[4]Tell little about individual
reviews
Review/reviewer/store reinforcementsFails on large number of
spam reviews with consistent ratings
Group spamming [2,3]No applicable on singleton reviews
Singleton reviews detection [7]*Finds suspicious hotels,
can’t find individual singleton spam
* [7] is a supervised method, and we have contrasting conclusion with theirs
The proposed method
•Recall the motivations of singleton reviews: boost the ratings in a short time and avoid being caught
•The results: in a short time, many reviewers wrote only one review with a very high rating
• The correlations between rating andThe correlations between rating andvolume of (singleton) reviews volume of (singleton) reviews is the key feature of singleton review is the key feature of singleton review spammingspamming
Detected burst of singleton spams
averagerating
numberof reviews
ratio ofsingleton reviews
a suspicious time window
The algorithm1.For each store do
A.split the whole period into small time windows
B.compute avg rating, total number of reviews, percentage of singleton reviews in each window
C.form a three dimension time series
1.detect windows with correlated burst patterns
2.for each detected window, repeat step A.-D. until window size becomes too small
55
The algorithm11
3322
4455
5544
1133
22
average rating: average rating: 22
review volume: review volume: 33
SR volume: 1/3SR volume: 1/3 average rating: average rating: 4.64.6
review volume: review volume: 55
SR volume: 5/5SR volume: 5/5 average rating: average rating: 22
review volume: review volume: 33
SR volume: 3/3SR volume: 3/3
sorted by sorted by posting posting time;time;
divided divided into into
groupsgroups
Multi-dimensional time series
the the correlated correlated
burstburst
Dataset• A snapshot of a review website*
• 408,469 reviews, 343,629 reviewers
• 310,499 reviewers (> 90%) wrote only one review
• 76% reviews are singleton reviews
• Focus on top 53 stores with over 1,000 reviews
* www.resellerratings.com
# reviewers
# reviews
Experimental results
•29 stores are regarded as suspicious by least 2 out of 3 human evaluators.
•The proposed algorithm labeled 39 stores as suspicious. ( recall = 75.86%, precision = 61.11%)
Case studies
time window size = 30 days
correlated bursts detected
Period with the detected correlated burst enlarged
time window size = 15 days
Volume of reviews: 57
Ratio of SRs: 61%
rating: 4.56
154
83%
4.79
pin-point the exact time and shape of the bursts
• Text features: ratio of reviews talking about “customer service/support”
• Hurry reviewers: wrote only one review at the same time of ID registration
• Human validation: read the reviews and found a reviewer disclosed being solicited for a 5 star review
Case studies (cont’)
Most of the later reviews are written by “Hurry Reviewers”
more than 80% of the singleton reviews are related to “customer service”
more than 80% of the singleton reviews are related to “customer service”
References1. Detecting Product Review Spammers using Rating Behaviors2. Finding Unusual Review Patterns Using Unexpected Rules3. Spotting Fake Reviewer Groups in Consumer Reviews4. Opinion spam and analysis5. Finding Deceptive Opinion Spam by Any Stretch of the Imagination6. Review Graph based Online Store Review Spammer Detection7. Merging Multiple Criteria to Identify Suspicious Reviews
the end
Examples All praisesPosted in a short timesay nothing about the
productsimilar ratingsRed flag words
mention the name a lot
Need to set up the animations for all these elements
Need to set up the animations for all these elements
Types of review spams
•Duplicate (easy, string matching)
•Advertisements
•Other easy-to-detect spams (all symbols, numbers, empty, etc.)
•untruthful (very hard, need machines to understand the intentions of the reviews)
Feature based methods
•Paradigms:
•Define features, training set and pick a classifier
•Keys to success: good features, large training data, and powerful classifier
Common but not fully investigated•Previous methods can not catch
them
•Each reviewer id has only one review
•Many features used by previous methods simply become meaningless
Traditional methods (cont.)
• [6] uses a graph to describe reinforcement relationships between entities: good / bad reviews influence their authors, who in turn influence the stores, which in turn influence its reviews.
reviews storesreviewers
Traditional methods (cont.)
reviews storesreviewers
{If these reviews are posted in a
short period with consistent ratings...
The store will be regarded as a good
one
Traditional methods
review featuresrating: 4, bag-of-words: {switch, nook, kindle, why,
what, learn}
reviewer features name: KKX; number of reviews: 1 average rating: 4
store/product features
Kindle average rating: 4/5 stars, price: $199
group spammingKKX wrote one review only, failed the frequency
test
what can what can you you
conclude conclude from these from these features?features?