From: KLOSURE: Closing in on open–ended patient questionnaires with text mining
Question | Topic | Classes | Method | Features | P (%) | R (%) | F (%) |
---|---|---|---|---|---|---|---|
Q1 | condition | 3 | MetaMap | N/A | 95.3 | 87.6 | 91.3 |
Q2 | treatment | 4 | MetaMap | N/A | 84.9 | 61.6 | 71.4 |
Q3 | changes | 3 | naive Bayes | 8 | 81.3 | 80.8 | 81.0 |
Q4 | confidence | 3 | best-first decision tree | 8 | 70.1 | 67.3 | 66.9 |
Q5 | stiffness | 3 | reduced error pruning tree | 8 | 85.3 | 79.6 | 75.6 |
Q6 | pain | 3 | complement naive Bayes | 10 | 62.8 | 58.3 | 59.0 |
Q7 | other symptoms | 3 | naive Bayes | 5 | 83.2 | 83.0 | 81.0 |
Q8 | daily activities | 3 | J48 pruned tree | 14 | 77.3 | 72.3 | 73.2 |
Q9 | other activities | 3 | random forest | 14 | 71.0 | 70.2 | 70.6 |
Q10 | other comments | 3 | Stanford Core NLP | N/A | 75.3 | 72.3 | 73.8 |