Rule-Ranking-Based Approximate Knowledge Interpolation With Directional Monotonicity