We define five feature sets based on different types of linguistic information, to gain further insight into the properties of each class. In particular, we are interested in the properties of event-related adjectives, for which we do not have a description in the linguistic literature. Table 8 summarizes the properties of the feature sets used for the present experiments.
Feature sets. From left to right, each column depicts, for each feature set, an identifier, a description of the type of information used, the total number of features, and one example feature. Feature set morph contains two categorical features that are transformed into 25 if binarization is applied; the remaining feature sets are numerical.
| Feature set . | Description . | # . | Example . | |
|---|---|---|---|---|
| morph | morphological (derivational) properties | 2 (25) | suffix | |
| func | syntactic function of the adjective | 4 | post-nom. modifier | |
| uni | uni-gram POS (1 word to left or to right) | 24 | −1noun | |
| bi | bi-gram POS (1 word to left and 1 to right) | 50 | −1noun+1adj | |
| theor | distributional cues of theoretical properties | 18 | gradable | |
| Total | 98 (121) |
| Feature set . | Description . | # . | Example . | |
|---|---|---|---|---|
| morph | morphological (derivational) properties | 2 (25) | suffix | |
| func | syntactic function of the adjective | 4 | post-nom. modifier | |
| uni | uni-gram POS (1 word to left or to right) | 24 | −1noun | |
| bi | bi-gram POS (1 word to left and 1 to right) | 50 | −1noun+1adj | |
| theor | distributional cues of theoretical properties | 18 | gradable | |
| Total | 98 (121) |