Table 1.
Unexpected word identification method | P | R | F1 |
---|---|---|---|
| |||
Baseline | 46.3 | 74.0 | 57.0 |
| |||
TF-IDF | 72.1 | 79.5 | 75.6 |
Log-odds | 70.5 | 79.5 | 74.7 |
Sum(TF-IDF, Log-odds) | 72.2 | 83.3 | 77.4 |
Max(TF-IDF, Log-odds) | 69.9 | 83.3 | 76.0 |
| |||
TF-IDF+WordNet | 83.8 | 80.5 | 82.1 |
Log-odds+WordNet | 82.1 | 83.1 | 82.6 |
Sum(TF-IDF, Log-odds)+WordNet | 84.2 | 83.1 | 83.7 |
Max(TF-IDF, Log-odds)+WordNet | 83.3 | 84.4 | 83.9 |
| |||
TF-IDF+WordNet+topic | 85.7 | 77.9 | 81.7 |
Log-odds+WordNet+topic | 83.8 | 80.5 | 82.1 |
Sum(TF-IDF, Log-odds)+WordNet+topic | 86.1 | 80.5 | 83.2 |
Max(TF-IDF, Log-odds)+WordNet+topic | 85.1 | 81.8 | 83.4 |