Skip to main content

Table 4 Five Way Cross-Validation Sentence Classification Results on Structured Abstracts: Using CRFs

From: Sentence retrieval for abstracts of randomized controlled trials

 

With no windowed features

With windowed features

 

Precision

Recall

F-score

Precision

Recall

F-score

1575 abs, 21.2 k sents

Accuracy = 86.30%

Accuracy = 87.99%

Aim

0.99

0.99

0.99

1.00

0.99

1.00

Method

0.85

0.84

0.85

0.87

0.85

0.86

Intervention

0.86

0.80

0.83

0.88

0.82

0.85

Results

0.79

0.84

0.82

0.82

0.87

0.84

Conclusion

0.93

0.92

0.93

0.94

0.93

0.93

1740 abs, 22.9 k sents

Accuracy = 95.17%

Accuracy = 95.10%

Aim

0.99

0.99

0.99

0.99

0.99

0.99

Method

0.97

0.97

0.97

0.96

0.97

0.97

Outcome Measure

0.90

0.86

0.88

0.90

0.86

0.88

Results

0.95

0.96

0.95

0.94

0.96

0.95

Conclusion

0.94

0.93

0.94

0.94

0.93

0.93

2280 abs, 29.8 k sents

Accuracy = 86.74%

Accuracy = 88.43%

Aim

0.99

0.99

0.99

0.99

0.99

0.99

Method

0.85

0.86

0.85

0.87

0.87

0.87

Participants

0.86

0.79

0.82

0.89

0.80

0.84

Results

0.81

0.84

0.83

0.83

0.87

0.85

Conclusion

0.93

0.93

0.93

0.94

0.93

0.94

  1. Sentence classification using CRFs into five classes, for each of the three classification problems. Results report using 15-fold cross validation for a system that uses no windowed features versus a system that uses windowed features.