Learning personalized individual semantics through the data of distributed linguistic preference relations: A two-stage method to support linguistic consensus reaching.