Research Article
A Subjective Expressions Extracting Method for Social Opinion Mining
Table 3
Measurement of ESE recognition model under different feature combinations.
| Feature combination | Binary Overlap | Proportional Overlap | Accuracy (%) | Recall (%) | F value (%) | Accuracy (%) | Recall (%) | F value (%) |
| +Pretrained word vectors | 82.83 | 48.79 | 61.24 | 67.11 | 43.14 | 52.27 | +Fine-tuning word vectors | 75.55 | 56.27 | 64.39 | 60.70 | 49.66 | 54.48 | +Fine-tuning word vectors + POS | 76.82 | 54.74 | 63.82 | 62.16 | 48.22 | 54.16 | +Fine-tuning word vectors + NER | 76.26 | 54.49 | 63.49 | 62.59 | 47.33 | 53.76 | +Fine-tuning word vectors + POS + NER | 75.56 | 57.52 | 65.15 | 60.80 | 50.85 | 55.07 |
|
|