Research Article
A Joint Model of Natural Language Understanding for Human-Computer Conversation in IoT
Table 2
Comparison with published results of joint models on the ATIS and SNIPS datasets.
| | Model | | | | Overall (Acc) | Intent (Acc) | Slot (F1) | Overall (Acc) | Intent (Acc) | Slot (F1) |
| | Joint Seq. [22] | 73.2 | 96.9 | 87.3 | 80.7 | 92.6 | 94.3 | | Attention-based RNN [24] | 74.1 | 96.7 | 87.8 | 78.9 | 91.1 | 94.2 | | Slot-gated [25] | 75.5 | 97.0 | 88.8 | 82.2 | 93.6 | 94.8 | | Capsule-NLU [28] | 80.9 | 97.3 | 91.8 | 83.4 | 95.0 | 95.2 | | SF-ID, SF-First [29] | 80.6 | 97.4 | 91.4 | 86.8 | 97.8 | 95.8 | | Bi-Model [30] | 83.8 | 97.2 | 93.5 | 85.7 | 96.4 | 95.5 | | Stack-Propagation [36] | 86.9 | 98.0 | 94.2 | 86.5 | 96.9 | 95.9 | | Bi-transfer (our model) | 85.9 | 98.3 | 95.5 | 88.0 | 98.7 | 96.0 |
|
|
