Wikidata is the new, large-scale knowledge base of the Wikimedia Foundation. Its knowledge is increasingly used within Wikipedia itself and various other kinds of information systems, imposing high demands on its integrity. Wikidata can be edited by anyone and, unfortunately, it frequently gets vandalized, exposing all information systems using it to the risk of spreading vandalized and falsified information. In this paper, we present a new machine learning-based approach to detect vandalism in Wikidata. We propose a set of 47 features that exploit both content and context information, and we report on 4 classifiers of increasing effectiveness tailored to this learning task. Our approach is evaluated on the recently published Wikidata Vandalism Corpus WDVC-2015 and it achieves an area under curve value of the receiver operating characteristic, ROC-AUC, of 0.991. It significantly outperforms the state of the art represented by the rule-based Wikidata Abuse Filter (0.865 ROC-AUC) and a prototypical vandalism detector recently introduced by Wikimedia within the Objective Revision Evaluation Service (0.859 ROC-AUC). http://www.heindorf.me/wdvd

1. Vandalism Detection in Wikidata
Stefan Heindorf1, Martin Potthast2, Benno Stein2, Gregor Engels1
CIKM 2016
October 25, 2016
1 2

2. Motivation
Vandalism Detection in Wikidata Stefan Heindorf 2

3. Motivation
Vandalism Detection in Wikidata Stefan Heindorf 2

4. Motivation
Vandalism Detection in Wikidata Stefan Heindorf 2

5. Motivation
Vandalism Detection in Wikidata Stefan Heindorf 2

6. Motivation
Vandalism Detection in Wikidata Stefan Heindorf 2

7. Motivation
Vandalism Detection in Wikidata Stefan Heindorf 2

8. Motivation
Vandalism Detection in Wikidata Stefan Heindorf 2

9. Motivation
Vandalism Detection in Wikidata Stefan Heindorf 2

10. 3Stefan HeindorfVandalism Detection in Wikidata

11. 3Stefan HeindorfVandalism Detection in Wikidata
Item head

12. 3Stefan HeindorfVandalism Detection in Wikidata
Item head
Item body

13. 3Stefan HeindorfVandalism Detection in Wikidata
Item head
Item body
Revisions

14. 3Stefan HeindorfVandalism Detection in Wikidata
(Feb 22, 2013)
(May 13, 2013)
(May 30, 2013)
Item head
Item body
Revisions

15. 3Stefan HeindorfVandalism Detection in Wikidata
(Feb 22, 2013)
(May 13, 2013)
(May 30, 2013)
Item head
Item body
Revisions

16. 3Stefan HeindorfVandalism Detection in Wikidata
(Feb 22, 2013)
(May 13, 2013)
(May 30, 2013)
Item head
Item body
Revisions

17. 3Stefan HeindorfVandalism Detection in Wikidata
Item head
Item body
Revisions

18. Why is it a problem?
4
Patrolling Reverting Warning Blocking Protecting
• Over 2 Mio manual edits per month
• A lot of tedious work
• Vandalism is not detected in time
Stefan HeindorfVandalism Detection in Wikidata

19. Research Question
How to detect damaging changes to
crowdsourced knowledge bases?
5Stefan HeindorfVandalism Detection in Wikidata

20. Our Approach
Vandalism Detection in Wikidata Stefan Heindorf 6

21. Our Approach
1. Label Dataset  Vandalism Corpus [SIGIR’15]
Vandalism Detection in Wikidata Stefan Heindorf 6

22. Our Approach
1. Label Dataset  Vandalism Corpus [SIGIR’15]
2. Study Vandalism Characteristics  47 Features
Vandalism Detection in Wikidata Stefan Heindorf 6

23. Our Approach
1. Label Dataset  Vandalism Corpus [SIGIR’15]
2. Study Vandalism Characteristics  47 Features
3. Experiment with ML  Multiple-Instance Learning
Vandalism Detection in Wikidata Stefan Heindorf 6

24. Our Approach
1. Label Dataset  Vandalism Corpus [SIGIR’15]
2. Study Vandalism Characteristics  47 Features
3. Experiment with ML  Multiple-Instance Learning
4. Compare with state of the art  2 Baselines
Vandalism Detection in Wikidata Stefan Heindorf 6

25. Corpus [SIGIR ’15]
Revisions over time
7

26. Corpus [SIGIR ’15]
Revisions over time
7Month

27. Corpus [SIGIR ’15]
Revisions over time
7Month

28. Corpus [SIGIR ’15]
Revisions over time
7Month
103,000 vandalism revisions

29. Corpus [SIGIR ’15]
Revisions over time
7Month
103,000 vandalism revisions
24 million manual revisions

30. Corpus [SIGIR ’15]
Revisions over time
7Month
103,000 vandalism revisions
24 million manual revisions
 0.4% vandalism

31. Corpus [SIGIR ’15]
Revisions over time
7
Item head
(1.3% vandalism)
Month
103,000 vandalism revisions
24 million manual revisions
 0.4% vandalism

32. Corpus [SIGIR ’15]
Revisions over time
7
Item head
(1.3% vandalism)
Item body
(0.2% vandalism)
Month
103,000 vandalism revisions
24 million manual revisions
 0.4% vandalism

33. Corpus [SIGIR ’15]
Revisions over time
7
Item head
(1.3% vandalism)
Item body
(0.2% vandalism)
Training
Month
103,000 vandalism revisions
24 million manual revisions
 0.4% vandalism

34. Corpus [SIGIR ’15]
Revisions over time
7
Item head
(1.3% vandalism)
Item body
(0.2% vandalism)
Training
Validation
Month
103,000 vandalism revisions
24 million manual revisions
 0.4% vandalism

35. Corpus [SIGIR ’15]
Revisions over time
7
Item head
(1.3% vandalism)
Item body
(0.2% vandalism)
Training
TestValidation
Month
103,000 vandalism revisions
24 million manual revisions
 0.4% vandalism

36. Content Features
11 Character features (e.g., lowerCaseRatio, digitRatio)
9 Word features (e.g., badWordRatio)
4 Sentence features (e.g., commentSitelinkSimilarity)
3 Statement features (e.g., propertyFrequency)
Context Features
10 User features (e.g., userCountry)
2 Item features (e.g., logItemFrequency)
8 Revision features (e.g., revisionTag, revisionLanguage)
Features (47 in total)
Stefan Heindorf 8Vandalism Detection in Wikidata

37. Features (47 in total)
Stefan Heindorf 8Vandalism Detection in Wikidata
revisionTag

38. Features (47 in total)
Stefan Heindorf 8Vandalism Detection in Wikidata
revisionTag Vand. Total Prob.
Rev. with tags 52 T 8,619 T 0.60%
By abuse filter 49 T 122 T 39.90%
By editing tools 3 T 8,496 T 0.03%
Rev. w/o tags 52 T 15,386 T 0.34%
revisionTag

39. Features (47 in total)
Stefan Heindorf 8Vandalism Detection in Wikidata
revisionTag Vand. Total Prob.
Rev. with tags 52 T 8,619 T 0.60%
By abuse filter 49 T 122 T 39.90%
By editing tools 3 T 8,496 T 0.03%
Rev. w/o tags 52 T 15,386 T 0.34%
revisionTag

40. Features (47 in total)
Stefan Heindorf 8Vandalism Detection in Wikidata
revisionTag Vand. Total Prob.
Rev. with tags 52 T 8,619 T 0.60%
By abuse filter 49 T 122 T 39.90%
By editing tools 3 T 8,496 T 0.03%
Rev. w/o tags 52 T 15,386 T 0.34%
revisionTag

41. Features (47 in total)
Stefan Heindorf 8Vandalism Detection in Wikidata
revisionTag Vand. Total Prob.
Rev. with tags 52 T 8,619 T 0.60%
By abuse filter 49 T 122 T 39.90%
By editing tools 3 T 8,496 T 0.03%
Rev. w/o tags 52 T 15,386 T 0.34%
revisionTag

42. Multiple-Instance Learning
Vandalism Detection in Wikidata Stefan Heindorf 9

43. Multiple-Instance Learning
• Observation: Vandalism seldom occurs in isolation
Vandalism Detection in Wikidata Stefan Heindorf 9

44. Multiple-Instance Learning
• Observation: Vandalism seldom occurs in isolation
Vandalism Detection in Wikidata Stefan Heindorf 9
22:35, 11 September 2013 184.19.64.111 (talk) . . (Changed English label: Barack Obama Aloha)
22:35, 11 September 2013 184.19.64.111 (talk) . . (Added English alias: Lulu:):):):):):):))
12:05, 11 September 2013 MatmaBot (talk | contribs) . . (Changed Polish description: imported

45. Multiple-Instance Learning
• Observation: Vandalism seldom occurs in isolation
Vandalism Detection in Wikidata Stefan Heindorf 9
22:35, 11 September 2013 184.19.64.111 (talk) . . (Changed English label: Barack Obama Aloha)
22:35, 11 September 2013 184.19.64.111 (talk) . . (Added English alias: Lulu:):):):):):):))
12:05, 11 September 2013 MatmaBot (talk | contribs) . . (Changed Polish description: imported

46. Multiple-Instance Learning
• Observation: Vandalism seldom occurs in isolation
Vandalism Detection in Wikidata Stefan Heindorf 9
22:35, 11 September 2013 184.19.64.111 (talk) . . (Changed English label: Barack Obama Aloha)
22:35, 11 September 2013 184.19.64.111 (talk) . . (Added English alias: Lulu:):):):):):):))
12:05, 11 September 2013 MatmaBot (talk | contribs) . . (Changed Polish description: imported
Session 1
Session 2

47. Multiple-Instance Learning
• Observation: Vandalism seldom occurs in isolation
• Idea: Apply Multiple-Instance Learning
Vandalism Detection in Wikidata Stefan Heindorf 9
22:35, 11 September 2013 184.19.64.111 (talk) . . (Changed English label: Barack Obama Aloha)
22:35, 11 September 2013 184.19.64.111 (talk) . . (Added English alias: Lulu:):):):):):):))
12:05, 11 September 2013 MatmaBot (talk | contribs) . . (Changed Polish description: imported
Session 1
Session 2

48. Multiple-Instance Learning
• Observation: Vandalism seldom occurs in isolation
• Idea: Apply Multiple-Instance Learning
Vandalism Detection in Wikidata Stefan Heindorf 9
22:35, 11 September 2013 184.19.64.111 (talk) . . (Changed English label: Barack Obama Aloha)
22:35, 11 September 2013 184.19.64.111 (talk) . . (Added English alias: Lulu:):):):):):):))
12:05, 11 September 2013 MatmaBot (talk | contribs) . . (Changed Polish description: imported
Session 1
Session 2

49. Multiple-Instance Learning
• Observation: Vandalism seldom occurs in isolation
• Idea: Apply Multiple-Instance Learning
Vandalism Detection in Wikidata Stefan Heindorf 9
22:35, 11 September 2013 184.19.64.111 (talk) . . (Changed English label: Barack Obama Aloha)
22:35, 11 September 2013 184.19.64.111 (talk) . . (Added English alias: Lulu:):):):):):):))
12:05, 11 September 2013 MatmaBot (talk | contribs) . . (Changed Polish description: imported
Session 1
Session 2

50. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
10Vandalism Detection in Wikidata
Stefan Heindorf

51. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
10Vandalism Detection in Wikidata
Stefan Heindorf

52. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10Vandalism Detection in Wikidata
Stefan Heindorf

53. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Precision
Recall
Vandalism Detection in Wikidata
Stefan Heindorf
Test Dataset (0.2% vandalism)

54. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Precision
Recall
Vandalism Detection in Wikidata
Stefan Heindorf
FILTER
Test Dataset (0.2% vandalism)

55. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Precision
Recall
Vandalism Detection in Wikidata
Stefan Heindorf
ORES
FILTER
Test Dataset (0.2% vandalism)

56. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Precision
Recall
Vandalism Detection in Wikidata
Stefan Heindorf
ORES
FILTER
Test Dataset (0.2% vandalism)

57. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Precision
Recall
Vandalism Detection in Wikidata
Stefan Heindorf
ORES
FILTER
Test Dataset (0.2% vandalism)

58. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Precision
Recall
Vandalism Detection in Wikidata
Stefan Heindorf
ORES
FILTER
PR-AUC: 0.491
ROC-AUC: 0.991
Test Dataset (0.2% vandalism)

59. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Precision
Recall
Vandalism Detection in Wikidata
Stefan Heindorf
 Detect and revert 30% vandalism
fully automatically
ORES
FILTER
Test Dataset (0.2% vandalism)

60. WDVD vs. Baselines
• WDVD (our approach)
Wikidata Vandalism Detector
• FILTER (baseline)
Wikidata Abuse Filter
• ORES (baseline)
Objective Revision Evaluation
Service
10
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Precision
Recall
Vandalism Detection in Wikidata
Stefan Heindorf
 Detect and revert 30% vandalism
fully automatically
ORES
FILTER
• Reduce workload by factor 10
(precision 2% instead of 0.2%)
 Still find 98.8% of all vandalism
Test Dataset (0.2% vandalism)

61. Conclusion and Outlook
Stefan Heindorf 11Vandalism Detection in Wikidata

62. Conclusion and Outlook
Conclusion
• Vandalism: Concentration on item heads (currently)
• Features: Content & Context
• Model: Multiple-Instance
• PR-AUC: 0.491
• ROC-AUC: 0.991
Stefan Heindorf 11Vandalism Detection in Wikidata

63. Conclusion and Outlook
Conclusion
• Vandalism: Concentration on item heads (currently)
• Features: Content & Context
• Model: Multiple-Instance
• PR-AUC: 0.491
• ROC-AUC: 0.991
Stefan Heindorf 11Vandalism Detection in Wikidata
Code + Data:
http://www.heindorf.me/
wdvd.html

64. Conclusion and Outlook
Conclusion
• Vandalism: Concentration on item heads (currently)
• Features: Content & Context
• Model: Multiple-Instance
• PR-AUC: 0.491
• ROC-AUC: 0.991
Outlook
• Goal: Better detection (on item bodies)
• Idea: Double-check with other sources
Stefan Heindorf 11Vandalism Detection in Wikidata
Code + Data:
http://www.heindorf.me/
wdvd.html

65. Conclusion and Outlook
Conclusion
• Vandalism: Concentration on item heads (currently)
• Features: Content & Context
• Model: Multiple-Instance
• PR-AUC: 0.491
• ROC-AUC: 0.991
Outlook
• Goal: Better detection (on item bodies)
• Idea: Double-check with other sources
Stefan Heindorf 11Vandalism Detection in Wikidata
Code + Data:
http://www.heindorf.me/
wdvd.html
Join the competition:
Vandalism Detection @WSDM Cup 2017
http://www.wsdm-cup-2017.org/

66. Conclusion and Outlook
Conclusion
• Vandalism: Concentration on item heads (currently)
• Features: Content & Context
• Model: Multiple-Instance
• PR-AUC: 0.491
• ROC-AUC: 0.991
Outlook
• Goal: Better detection (on item bodies)
• Idea: Double-check with other sources
Acknowledgement
• German Research Foundation (DFG)
• SIGIR Student Travel Grant
Stefan Heindorf 11Vandalism Detection in Wikidata
Code + Data:
http://www.heindorf.me/
wdvd.html
Join the competition:
Vandalism Detection @WSDM Cup 2017
http://www.wsdm-cup-2017.org/

67. Conclusion and Outlook
Conclusion
• Vandalism: Concentration on item heads (currently)
• Features: Content & Context
• Model: Multiple-Instance
• PR-AUC: 0.491
• ROC-AUC: 0.991
Outlook
• Goal: Better detection (on item bodies)
• Idea: Double-check with other sources
Acknowledgement
• German Research Foundation (DFG)
• SIGIR Student Travel Grant
Stefan Heindorf 11Vandalism Detection in Wikidata
Code + Data:
http://www.heindorf.me/
wdvd.html
Join the competition:
Vandalism Detection @WSDM Cup 2017
http://www.wsdm-cup-2017.org/
Thank you!