Coherence evaluation for Dialogue Generation

I proposed a variation of a new metric and a new pipeline to evaluate the systems. With the main objective to tackle the drawbacks of the actual metrics without incorporating humans in the loop