A. D. Duque, L. L. Santos, D. Macêdo, C. Zanchettin, "Squeezed Very Deep Convolutional Networks for Text Classification".
| Dataset | Classes | Train samples | Test samples | source |
|---|---|---|---|---|
| AG’s News | 4 | 120 000 | 7 600 | link |
| Yelp Review Polarity | 2 | 560 000 | 38 000 | link |
| Yelp Review Full | 5 | 650 000 | 50 000 | link |
Thanks to @ArdalanM for the VDCNN implentation used as baseline (here)
Execute the run.sh file through the command:
- bash ./run.sh
| SVDCNN | VDCNN | Char-CNN | |
|---|---|---|---|
| 6 layers | --- | --- | 43.25 |
| 9 layers | 2.80 | 54.75 | --- |
| 17 layers | 5.52 | 62.74 | --- |
| 29 layers | 6.03 | 64.16 | --- |
| SVDCNN | VDCNN | Char-CNN | |
|---|---|---|---|
| 6 layers | --- | --- | 92.36 |
| 9 layers | 90.13 | 90.83 | --- |
| 17 layers | 90.43 | 91.12 | --- |
| 29 layers | 90.55 | 91.27 | --- |
| SVDCNN | VDCNN | Char-CNN | |
|---|---|---|---|
| 6 layers | --- | --- | 95.64 |
| 9 layers | 94.99 | 95.12 | --- |
| 17 layers | 95.04 | 95.50 | --- |
| 29 layers | 95.26 | 95.72 | --- |
| SVDCNN | VDCNN | Char-CNN | |
|---|---|---|---|
| 6 layers | --- | --- | 62.05 |
| 9 layers | 61.97 | 63.27 | --- |
| 17 layers | 63.00 | 63.93 | --- |
| 29 layers | 63.20 | 64.26 | --- |