Table 7 Comparison of entity detection performance on testing dataset (in percent). SQ and LQ refer to SimpleQuestions and LC-QuAD 2.0 dataset, respectively
Approach and the dataset(s) used | Acc. | P | R | F1 |
|---|---|---|---|---|
A. Tested on SQ: | ||||
Dai et al. [7] | 75.7 | - | - | - |
He and Golub [8] | 96.8 | - | - | - |
Chao and Li [9] | 82.2 | - | - | - |
Lukovnikov et al. [10] | 82.7 | - | - | - |
Azmy et al. [16] | - | - | - | 90.3 |
Cui et al. [15] | - | 97.4 | 96.1 | 96.1 |
Our approach trained on SQ | 99.15 | 99.25 | 99.15 | 99.17 |
Our approach trained on SQ, LQ, and QALD | 97.1 | 97.9 | 97.1 | 97.4 |
B. Tested on LQ: | ||||
Evseev and Arkhipov [11] | - | - | - | 87 |
Our approach trained on LQ | 74.09 | 75.72 | 74.09 | 73.87 |
Our approach trained on SQ, LQ, and QALD | 97.4 | 98 | 97.4 | 97.6 |
C. Tested on QALD | ||||
Hu, Zou, Yu et al. [13] (QALD 6) | 92 | – | – | – |
Bakhshi et al. [14] (tested on QALD 9) | 78 | – | – | – |
*Kuo and Lu [12] (tested on QALD 7) | 91 | – | – | – |
Our approach: (trained on SQ, LQ, and QALD) | ||||
Tested on QALD-1 | 95.4 | – | – | – |
Tested on QALD-2 | 100 | – | – | – |
Tested on QALD-3 | 100 | – | – | – |
Tested on QALD-4 | 97.5 | – | – | – |
Tested on QALD-5 | 95.1 | – | – | – |
Tested on QALD-6 | 96.4 | – | – | – |
Tested on QALD-7 | 95.4 | – | – | – |
Tested on QALD-8 | 88.2 | – | – | – |
Tested on QALD-9 | 96 | – | – | – |