Research Article
Multigranularity Pruning Model for Subject Recognition Task under Knowledge Base Question Answering When General Models Fail
Table 1
Experimental results for subject recognition.
| | Method | SQ | Dataset I | Dataset II | WSP | Dataset III | Dataset IV | WQ |
| | MGPM | 89.5 | 92.4 | 46.0 | 52.8 | 61.6 | 26.1 | 52.1 | | BERT-CRF by bert4keras | 90.8 | 97.0 | 0 | 60.0 | 79.8 | 0 | 58.4 | | +MGPM | 94.4 (3.6) | 98.4 (1.4) | 36.0 | 68.6 (8.6) | 87.3 (7.5) | 11.8 | 63.7 (5.3) | | EGP | 89.4 | 95.4 | 0 | 62.2 | 82.7 | 0 | 60.4 | | +MGPM | 94.1 (4.7) | 98.0 (2.6) | 37.0 | 70.3 (8.1) | 89.5 (6.8) | 11.8 | 65.2 (4.8) |
|
|
Bold values represent the best-performance.
|
