| D09-1049 | have experimented with additional | feature combinations | , with no improvement in results |
| D09-1056 | shows this correlation for several | features combinations | . This is an indication that |
| D08-1049 | adding character n-grams to other | feature combinations | always gives significant improvements |
| C86-1140 | nucleus class . Of all possible | feature combinations | , the one with the highest consistency |
| D09-1160 | classifiers and their ways to consider | feature combinations | . In classification-based NLP |
| D09-1107 | extends easily to cover non-linear | feature combinations | . The reason for using amoeba |
| D09-1056 | coreferent pairs lower . In the case of | feature combinations | we can study them by training |
| D08-1024 | Marton and Resnik 's XP = and XP ' | feature combinations | , respectively . Fine-grained |
| C04-1088 | classification accuracy below the other | feature combinations | . In combination , though , these |
| D09-1056 | next section . 4.3 Analysis of | feature combinations | Up to now we have analysed the |
| D08-1087 | development set are starred . 4.5 | Feature Combination | Unlike most previous work , n-gram |
| D09-1111 | opportunity to improve through system or | feature combination | : an oracle that perfectly selects |
| C04-1166 | Passages : Exploring Linguistic | Feature Combinations | via Machine Learning . In Proceedings |
| D08-1024 | many larger , potentially better | feature combinations | remained unexplored . Moreover |
| D09-1056 | features . First , we use each | feature combinations | as the input for a Machine Learning |
| D08-1024 | features and manually selected | feature combinations | each in a separate model . Although |
| C90-3030 | there a reading with any of the | feature combinations | defining finite verbs . The condition |
| D08-1087 | roughly half for a variety of | feature combinations | . However , all improvements |
| C04-1190 | by exploiting the similarity to | feature combinations | that have been seen . However |
| D09-1032 | ) = w N ! n1!n2 ! ... nr ! 4.4 | Feature combination | using linear regression We also |
