Table 3.

Classification results obtained over the GXD-caption dataset using different set of features. AB indicates using features form titles/abstracts only. CAP indicates using features from captions alone. AB&CAP indicates using features from both captions and titles/abstracts. NB denotes Naáve Bayes; RF denotes Random Forest classifier. Standard deviation is shown in parentheses. The best performance in each metric is indicated using Boldface.

Classifiers	Precision	Recall	F-measure	Accuracy	Utility-10	Utility-20
NB_AB	.874(.023)	.656(.027)	.749(.022)	.783(.017)	.647(.027)	.652(.027)
RF_AB	.779(.016)	.768(.024)	.773(.017)	.802(.015)	.746(.024)	.757(.024)
NB_CAP	.831(.018)	.766(.024)	.797(.019)	.808(.017)	.751(.024)	.758(.024)
RF_CAP	.855(.019)	.758(.017)	.804(.015)	.817(.014)	.745(.018)	.752(.018)
NB_AB&CAP	.877(.018)	.816(.018)	.846(.014)	.853(.013)	.805(.018)	.810(.018)
RF_AB&CAP	.876(.019)	.829(.015)	.852(.013)	.858(.012)	.817(.015)	.823(.015)