Assessing the calibration of naive Bayes' posterior estimatesProbabilistic combination of text classifiers using reliability indicators: Models and results