Crowdsourcing using Mechanical Turk

Quality Management and Scalability

Panos Ipeirotis – New York University

“A Computer Scientist in a Business

School”

http://behind-the-enemy-lines.blogspot

.com/

Email: panos@nyu.edu

“A Computer Scientist in a Business

School”

http://behind-the-enemy-lines.blogspot

.com/

Email: panos@nyu.edu

Panos Ipeirotis - Introduction

New York University, Stern School of Business

4

5

Example: Build an “Adult Web Site” Classifier

Need a large number of hand-labeled sites Get people to look at sites and classify them as:G (general audience) PG (parental guidance) R (restricted) X

(porn)

Cost/Speed Statistics Undergrad intern: 200 websites/hr, cost:

$15/hr

Cost/Speed Statistics Undergrad intern: 200 websites/hr, cost:

$15/hr

Amazon Mechanical Turk: Paid Crowdsourcing

Example: Build an “Adult Web Site” Classifier

Need a large number of hand-labeled sites Get people to look at sites and classify them as:G (general audience) PG (parental guidance) R (restricted) X

(porn)

Cost/Speed Statistics Undergrad intern: 200 websites/hr, cost:

$15/hr MTurk: 2500 websites/hr, cost: $12/hr

Cost/Speed Statistics Undergrad intern: 200 websites/hr, cost:

$15/hr MTurk: 2500 websites/hr, cost: $12/hr

Bad news: Spammers!

Worker ATAMRO447HWJQ

labeled X (porn) sites as G (general

audience)

Worker ATAMRO447HWJQ

labeled X (porn) sites as G (general

audience)

Improve Data Quality through Repeated Labeling

Get multiple, redundant labels using multiple workers Pick the correct label based on majority vote

Probability of correctness increases with number of workers

Probability of correctness increases with quality of workers

1 worker

70%

correct

1 worker

70%

correct

11 workers

93%

correct

11 workers

93%

correct

11-vote Statistics MTurk: 227 websites/hr, cost: $12/hr Undergrad: 200 websites/hr, cost:

$15/hr

11-vote Statistics MTurk: 227 websites/hr, cost: $12/hr Undergrad: 200 websites/hr, cost:

$15/hr

Single Vote Statistics MTurk: 2500 websites/hr, cost: $12/hr Undergrad: 200 websites/hr, cost:

$15/hr

Single Vote Statistics MTurk: 2500 websites/hr, cost: $12/hr Undergrad: 200 websites/hr, cost:

$15/hr

But Majority Voting is Expensive

Using redundant votes, we can infer worker quality

Look at our spammer friend ATAMRO447HWJQ together with other 9 workers

Our “friend” ATAMRO447HWJQ mainly marked sites as G.Obviously a spammer…

We can compute error rates for each worker

Error rates for ATAMRO447HWJQ P[X → X]=9.847% P[X → G]=90.153% P[G → X]=0.053% P[G → G]=99.947%

Rejecting spammers and BenefitsRandom answers error rate = 50%

Average error rate for ATAMRO447HWJQ: 45.2% P[X → X]=9.847% P[X → G]=90.153% P[G → X]=0.053% P[G → G]=99.947%

Action: REJECT and BLOCK

Results: Over time you block all spammers Spammers learn to avoid your HITS You can decrease redundancy, as quality of workers is higher

After rejecting spammers, quality goes up Spam keeps quality down Without spam, workers are of higher quality Need less redundancy for same quality Same quality of results for lower cost

With spam

1 worker

70%

correct

With spam

1 worker

70%

correct

With spam

11 workers

93%

correct

With spam

11 workers

93%

correct

Without

spam

1 worker

80% correct

Without

spam

1 worker

80% correct

Without

spam

5 workers

94% correct

Without

spam

5 workers

94% correct

Correcting biases

Classifying sites as G, PG, R, X Sometimes workers are careful but biased

Classifies G → P and P → R Average error rate for ATLJIK76YH1TF: too high

Is she a spammer?Is she a spammer?

Error Rates for CEO of AdSafe

P[G → G]=20.0% P[G → P]=80.0%P[G → R]=0.0% P[G → X]=0.0%P[P → G]=0.0% P[P → P]=0.0% P[P → R]=100.0% P[P → X]=0.0%P[R → G]=0.0% P[R → P]=0.0% P[R → R]=100.0% P[R → X]=0.0%P[X → G]=0.0% P[X → P]=0.0% P[X → R]=0.0% P[X → X]=100.0%

Error Rates for CEO of AdSafe

P[G → G]=20.0% P[G → P]=80.0%P[G → R]=0.0% P[G → X]=0.0%P[P → G]=0.0% P[P → P]=0.0% P[P → R]=100.0% P[P → X]=0.0%P[R → G]=0.0% P[R → P]=0.0% P[R → R]=100.0% P[R → X]=0.0%P[X → G]=0.0% P[X → P]=0.0% P[X → R]=0.0% P[X → X]=100.0%

Correcting biases

For ATLJIK76YH1TF, we simply need to “reverse the errors” (technical details omitted) and separate error and bias

True error-rate ~ 9%

Error Rates for Worker: ATLJIK76YH1TF

P[G → G]=20.0% P[G → P]=80.0%P[G → R]=0.0% P[G → X]=0.0%P[P → G]=0.0% P[P → P]=0.0% P[P → R]=100.0% P[P → X]=0.0%P[R → G]=0.0% P[R → P]=0.0% P[R → R]=100.0% P[R → X]=0.0%P[X → G]=0.0% P[X → P]=0.0% P[X → R]=0.0% P[X → X]=100.0%

Error Rates for Worker: ATLJIK76YH1TF

P[G → G]=20.0% P[G → P]=80.0%P[G → R]=0.0% P[G → X]=0.0%P[P → G]=0.0% P[P → P]=0.0% P[P → R]=100.0% P[P → X]=0.0%P[R → G]=0.0% P[R → P]=0.0% P[R → R]=100.0% P[R → X]=0.0%P[X → G]=0.0% P[X → P]=0.0% P[X → R]=0.0% P[X → X]=100.0%

Too much theory?

Demo and Open source implementation available at:

http://qmturk.appspot.com Input:

– Labels from Mechanical Turk– Cost of incorrect labelings (e.g., XG costlier than GX)

Output: – Corrected labels– Worker error rates– Ranking of workers according to their quality

Beta version, more improvements to come! Suggestions and collaborations welcomed!

Scaling Crowdsourcing: Use Machine Learning Human labor is expensive, even when paying cents Need to scale crowdsourcing

Basic idea: Build a machine learning model and use it instead of humans

Data from existing

crowdsourced answers

Data from existing

crowdsourced answers

New CaseNew Case Automatic Model

(through machine

learning)

Automatic Model

(through machine

learning)

Automatic

Answer

Automatic

Answer

20

Tradeoffs for Automatic Models: Effect of Noise

Get more data Improve model accuracy Improve data quality Improve classification

Example Case: Porn or not?

40

50

60

70

80

90

100

1 20 40 60 80 100120140160180200220240260280300

Number of examples (Mushroom)

Acc

ura

cy

Data Quality = 50%

Data Quality = 60%

Data Quality = 80%

Data Quality = 100%

Confide

ntConfide

nt

Automatic Model

(through machine

learning)

Automatic Model

(through machine

learning)

Scaling Crowdsourcing: Iterative training

Use machine when confident, humans otherwise

Retrain with new human input → improve model → reduce need for humans

Get human(s)

to answer

Get human(s)

to answer

New CaseNew Case

Not confident

Not confident

Automatic

Answer

Automatic

Answer

Data from existing

crowdsourced answers

Data from existing

crowdsourced answers

22

Tradeoffs for Automatic Models: Effect of Noise

Get more data Improve model accuracy Improve data quality Improve classification

Example Case: Porn or not?

40

50

60

70

80

90

100

1 20 40 60 80 100120140160180200220240260280300

Number of examples (Mushroom)

Acc

ura

cy

Data Quality = 50%

Data Quality = 60%

Data Quality = 80%

Data Quality = 100%

Not confident

Not confident

Confiden

tConfiden

t

Automatic Model

(through machine

learning)

Automatic Model

(through machine

learning)

Scaling Crowdsourcing: Iterative training, with noise

Use machine when confident, humans otherwise Ask as many humans as necessary to ensure quality

Get human(s)

to answer

Get human(s)

to answer

New CaseNew Case

Automatic

Answer

Automatic

Answer

Confident for quality?

Not confident

for quality?

Data from existing

crowdsourced answers

Data from existing

crowdsourced answers

Thank you!

Questions?

“A Computer Scientist in a Business School”

http://behind-the-enemy-lines.blogspot.com

/

Email: panos@nyu.edu

“A Computer Scientist in a Business School”

http://behind-the-enemy-lines.blogspot.com

/

Email: panos@nyu.edu