The accuracies for the simple classifiers are shown in Table 10. Part A of the table lists the results for each of the binary decisions (qualitative/non-qualitative, event/non-event, relational/non-relational). The accuracy for each decision is computed independently. For instance, a qualitative-event adjective is judged correct within the qualitative class iff the decision is qualitative; correct within the event class iff the decision is event; and correct within the relational class iff the decision is non-relational.
Second model: Results with simple classifiers using different feature sets. The frequency baseline (first row) is marked in italics. The last row, headed by all, shows the accuracy obtained when using all features together for tree construction. The remaining rows follow the nomenclature in Table 8; a FS subscript indicates that automatic feature selection is used as explained in Section 4.2. For each feature set, we record the mean and the standard deviation (marked by ±) of the accuracies. Best and second best results are boldfaced. Significant improvements over the baseline are marked as follows: *p < 0.05; **p < 0.01; ***p < 0.001.
. | A: Per-class accuracy . | B: Overall accuracy . | |||
|---|---|---|---|---|---|
. | Qualitative . | Event . | Relational . | Full . | Partial . |
| baseline | 65.2 ± 11.1 | 76.2 ± 9.9 | 71.9 ± 9.6 | 51.0 ± 0.0 | 65.2 ± 0.0 |
| morph | 68.2 ± 11.1 | 87.3** ± 6.3 | 85.2*** ± 7.2 | 59.9*** ± 2.2 | 84.7*** ± 0.7 |
| morphFS | 72.5* ± 7.9 | 89.1** ± 6.0 | 84.2*** ± 7.5 | 60.6*** ± 1.3 | 87.8*** ± 0.4 |
| func | 75.1** ± 9.0 | 76.1 ± 9.8 | 82.8** ± 7.5 | 56.0*** ± 1.9 | 80.6*** ± 1.8 |
| uni | 64.2 ± 10.8 | 68.4 ± 12.0 | 82.1** ± 9.0 | 42.8 ± 2.7 | 74.8*** ± 2.6 |
| uniFS | 66.0 ± 9.3 | 75.1 ± 10.6 | 82.2** ± 7.5 | 52.9 ± 1.9 | 77.0*** ± 2.0 |
| bi | 63.8 ± 9.9 | 66.2 ± 9.8 | 78.2* ± 8.2 | 46.1 ± 2.3 | 77.8*** ± 1.8 |
| biFS | 67.4 ± 10.6 | 72.3 ± 10.2 | 83.0*** ± 8.3 | 52.3 ± 1.7 | 76.7*** ± 1.0 |
| theor | 71.8 ± 10.0 | 74.1 ± 9.9 | 86.4*** ± 7.6 | 54.8*** ± 1.7 | 81.8*** ± 1.8 |
| all | 75.5** ± 9.0 | 86.5** ± 6.4 | 86.0*** ± 6.5 | 62.5*** ± 2.5 | 87.6*** ± 2.5 |
. | A: Per-class accuracy . | B: Overall accuracy . | |||
|---|---|---|---|---|---|
. | Qualitative . | Event . | Relational . | Full . | Partial . |
| baseline | 65.2 ± 11.1 | 76.2 ± 9.9 | 71.9 ± 9.6 | 51.0 ± 0.0 | 65.2 ± 0.0 |
| morph | 68.2 ± 11.1 | 87.3** ± 6.3 | 85.2*** ± 7.2 | 59.9*** ± 2.2 | 84.7*** ± 0.7 |
| morphFS | 72.5* ± 7.9 | 89.1** ± 6.0 | 84.2*** ± 7.5 | 60.6*** ± 1.3 | 87.8*** ± 0.4 |
| func | 75.1** ± 9.0 | 76.1 ± 9.8 | 82.8** ± 7.5 | 56.0*** ± 1.9 | 80.6*** ± 1.8 |
| uni | 64.2 ± 10.8 | 68.4 ± 12.0 | 82.1** ± 9.0 | 42.8 ± 2.7 | 74.8*** ± 2.6 |
| uniFS | 66.0 ± 9.3 | 75.1 ± 10.6 | 82.2** ± 7.5 | 52.9 ± 1.9 | 77.0*** ± 2.0 |
| bi | 63.8 ± 9.9 | 66.2 ± 9.8 | 78.2* ± 8.2 | 46.1 ± 2.3 | 77.8*** ± 1.8 |
| biFS | 67.4 ± 10.6 | 72.3 ± 10.2 | 83.0*** ± 8.3 | 52.3 ± 1.7 | 76.7*** ± 1.0 |
| theor | 71.8 ± 10.0 | 74.1 ± 9.9 | 86.4*** ± 7.6 | 54.8*** ± 1.7 | 81.8*** ± 1.8 |
| all | 75.5** ± 9.0 | 86.5** ± 6.4 | 86.0*** ± 6.5 | 62.5*** ± 2.5 | 87.6*** ± 2.5 |