Computing and Evaluating Syntactic Complexity Features for Automated Scoring of Spontaneous Non-Native Speech