Table 1 Datasets characteristics.
From: A deep sentiment model combining ALBERT-driven context and EHO-optimized architecture
Dataset | Train (Pos/Neu/Neg) | Test (Pos/Neu/Neg) | Classes | Text length | Domain |
|---|---|---|---|---|---|
1561 / 3127 / 1560 | 173 / 346 / 173 | 3 | Short | Social media trends | |
Rest14 | 2164 / 637 / 807 | 728 / 196 / 196 | 3 | Medium–long | Restaurant reviews |
Rest15 | 912 / 36 / 256 | 326 / 34 / 182 | 3 | Medium–long | Restaurant reviews |
Rest16 | 1240 / 69 / 439 | 469 / 30 / 117 | 3 | Medium–long | Restaurant reviews |
Lap14 | 994 / 464 / 870 | 341 / 169 / 128 | 3 | Medium–long | Electronics reviews |
SST-5 | 2322 / 1624 / 2218 Very neg:1092 Very pos: 1288 | 510 / 389 / 633 Very neg:279 Very pos:399 | 5 | Variable | General sentiments |
