Research Article
A BERT-Based Aspect-Level Sentiment Analysis Algorithm for Cross-Domain Text
Table 3
Results of accuracy and F1.
| Source | Target | Accuracy | F1 value | NO | YES | NO | YES |
| B | D | 0.778 | 0.822 | 0.772 | 0.814 | E | 0.753 | 0.785 | 0.762 | 0.796 | K | 0.768 | 0.821 | 0.774 | 0.835 | V | 0.815 | 0.823 | 0.809 | 0.812 | B | 0.755 | 0.835 | 0.752 | 0.842 |
| D | E | 0.753 | 0.812 | 0.760 | 0.811 | K | 0.784 | 0.814 | 0.764 | 0.825 | V | 0.878 | 0.917 | 0.910 | 0.921 | B | 0.752 | 0.801 | 0.750 | 0.821 | D | 0.754 | 0.801 | 0.734 | 0.824 |
| E | K | 0.815 | 0.875 | 0.812 | 0.867 | V | 0.756 | 0.792 | 0.784 | 0.815 | B | 0.712 | 0.784 | 0.701 | 0.765 | D | 0.754 | 0.805 | 0.762 | 0.801 | E | 0.768 | 0.822 | 0.732 | 0.821 | V | 0.815 | 0.845 | 0.810 | 0.848 | B | 0.855 | 0.875 | 0.840 | 0.867 | D | 0.835 | 0.869 | 0.857 | 0.869 | E | 0.824 | 0.854 | 0.834 | 0.868 | K | 0.850 | 0.889 | 0.846 | 0.870 |
|
|