Research Article
Effective Preprocessing and Normalization Techniques for COVID-19 Twitter Streams with POS Tagging via Lightweight Hidden Markov Model
Table 13
Overall plotting parameters for ROC utilized for tweet preprocessing.
| Plotting parameters | Investigating OOV words in tweets | Tweet normalization and preprocessing | Evaluated POS tagging for tweets |
| Number of cases | 2108 | 450 | 950 | Number of correct | 1907 | 412 | 702 | Accuracy | 77.5% | 84% | 73.9% | Sensitivity | 75% | 88% | 68.6% | Specificity | 80% | 80% | 79.2% | Positive cases missed | 107 | 77 | 149 | Negative cases missed | 94 | 53 | 99 | Fitted ROC curve | 0.85 | 0.90 | 0.79 |
|
|