Research Article
IARNN-Based Semantic-Containing Double-Level Embedding Bi-LSTM for Question-and-Answer Matching
Table 6
Different performance measures on the test set.
| ā | Model | F1 (%) | Recall (%) | Top-2 (%) | Top-3 (%) |
| 1 | BM25 | 49.02 | 53.27 | 49.62 | 52.73 | 2 | Multiscale CNN | 62.15 | 59.83 | 67.85 | 69.53 | 3 | Attentive pooling | 73.18 | 73.86 | 75.97 | 78.21 | 4 | DMN | 74.74 | 75.12 | 77.42 | 80.93 | 5 | ESIM + ELMo | 77.54 | 78.26 | 79.27 | 80.31 | 6 | Multiview | 76.09 | 76.83 | 78.14 | 79.83 | 7 | CapsNet | 78.17 | 78.82 | 80.61 | 83.53 | 8 | SCDE-Bi-LSTM | 79.83 | 80.52 | 82.13 | 84.73 |
|
|