Decision tree construction is a popular approach in data mining and machine learning, and some variants of decision tree algorithms have been proposed to deal with different types of data. In this paper, we present a bipolar interpretation of fuzzy decision trees. With the interpretation, various types of decision trees can be represented in a unified form. The edges of a fuzzy decision tree are labeled by fuzzy decision logic formulas and the nodes are split according to the satisfaction of these formulas in the data records. We present a construction algorithm for general fuzzy decision trees and show its application to different types of training data.