Research Article
Automatic Approach of Sentiment Lexicon Generation for Mobile Shopping Reviews
Table 5
F1-measure classification results for shopping domain reviews (NB).
| Method | Chinese | English | Hotel | Cloth | Fruit | Books | Kitchen | DVD | Electronics |
| No lexicon | | | | | | | | BOW | 81.9 | 86.4 | 88.8 | 74.0 | 79.3 | 77.6 | 77.0 | W2V | 61.2 | 68.2 | 69.8 | 61.5 | 67.1 | 63.8 | 61.7 | General lexicon | | | | | | | | DUTIR | 71.5 | 75.0 | 73.8 | - | - | - | - | SentiWordNet | - | - | - | 72.3 | 79.0 | 76.8 | 76.5 | Only 3 | 68.8 | 70.7 | 70.2 | 65.1 | 62.3 | 68.4 | 61.9 | DUTIR+3 | 72.6 | 74.8 | 74.2 | - | - | - | - | SentiWordNet+3 | - | - | - | 74.9 | 79.8 | 78.5 | 77.9 | Domain-specific lexicon | | | | | | | | DS | 80.3 | 84.9 | 86.8 | 77.7 | 81.1 | 79.5 | 79.0 | Only | 70.3 | 71.2 | 71.5 | 61.6 | 64.9 | 62.7 | 63.2 | DS+ | 83.6 | 87.6 | 89.5 | 78.9 | 82.5 | 80.6 | 80.5 |
|
|