Corpus | Precision | Recall | F-M |
---|
Converted LLL
| 68.5 | 68.5 |
68.5
|
LLL
| 79.3 | 85.1 |
82.1
|
- Our walk-weighted subsequence kernel was trained on automatic-parsed LLL training data and then tested on automatic-parsed LLL test data. The result was evaluated by the LLL scoring server. It was compared with the result over clean LLL training/test set.