E14-1043 nonlinear methods for the impact score prediction task given the multimodality
P14-1144 Features Our baseline system for score prediction employs various features based
P98-1032 for each test question . Final score prediction for cross-validation uses the
P98-1032 human rater scores and e-rater score predictions . In accordance with human interrater
P98-1032 extracted and used for machine-based score prediction of essay responses . These three
P14-1123 more likely to provide better score predictions than VSM does . Feature design
P14-1144 44 105 230 443 4 Score Prediction In this section , we describe
P15-1053 the distance between a system 's score predictions and the annotator-assigned scores
P14-1144 the distance between a system 's score predictions and the annotator-assigned scores
P13-1026 thesis clarity errors . Because our score prediction system uses the same feature
P15-1053 cross-validation results on argument strength score prediction are shown in Table 6 . The first
E95-1026 information and is able to provide scored predictions in a fast and efficient way ,
P98-1032 Electronic Essay Rater ( e-rater ) score predictions and human rater scores ranged
P14-1144 cross-validation results on prompt adherence score prediction are shown in Table 4 . On the
P98-1032 features for each test question . Score prediction for cross-validation sets is
J12-4003 for inexact argument bounds , we scored predictions using the Dice coefficient ,
P15-1053 scores for argument strength . 4 Score Prediction We cast the task of predicting
P10-1160 argument position unfilled . We scored predictions using the Dice coefficient ,
D10-1023 differing lengths may be useful for score prediction , we create a binary presence
P98-1032 set identified during training . Score prediction accuracy is determined by measuring
hide detail