Skip to main content

Table 3 Comparative table of deduplication results following experts and Deduklick analysis

From: Reducing systematic review burden using Deduklick: a novel, automated, reliable, and explainable deduplication algorithm to foster medical research

Dataset

Type

ET s

True + 

True − 

False + 

False − 

Recall

Precision

F1

Sustain. food

Experts

4200

3157

4435

0

3

99.91%

100.00%

99.95%

Deduklick

49

3148

4435

0

12

99.62%

100.00%

99.81%

Healthy aging

Experts

4200

10,356

7853

6

99

99.05%

99.94%

99.50%

Deduklick

109

10,394

7859

0

61

99.42%

100.00%

99.71%

Healthy lifestyle

Experts

4200

5530

7888

0

104

98.15%

100.00%

99.07%

Deduklick

92

5592

7888

0

42

99.25%

100.00%

99.63%

Menopause onset

Experts

4200

3776

4227

2

52

98.64%

99.95%

99.29%

Deduklick

24

3814

4229

0

14

99.64%

100.00%

99.82%

Hypertension

Experts

4200

4546

9112

2

364

92.59%

99.96%

96.13%

Deduklick

106

4922

9114

0

5

99.90%

100.00%

99.95%

e3_gsm

Experts

4200

406

1223

1

46

89.82%

99.75%

94.53%

Deduklick

19

447

1224

0

5

98.89%

100.00%

99.44%

Jugular

Experts

4200

49

1236

0

109

31.01%

100.00%

47.34%

Deduklick

29

159

1236

0

1

99.38%

100.00%

99.69%

Clinical trials

Experts

4200

30

15

0

0

100.00%

100.00%

100.00%

Deduklick

2

30

15

0

0

100.00%

100.00%

100.00%

Averages

Experts

4200

3481.3

4498.6

1.4

97.1

88.65%

99.95%

91.98%

Deduklick

54

3563.3

4500

0

17.5

99.51%

100.00%

99.75%