Here, the answer key is very similar to the "semantic answer" of a tutoring assessment type in AutoTutor Lite. For each of your questions (good, average, poor, etc.) you should have a list of key words that are required to fully answer the question. You can use regular expressions to input your answer keys. 

These keys are what AutoTutor Lite will compare your users' input to. Select the threshold, or the % coverage of the keys in the user input, to trigger the feedback from your tutor. Here, the user must use about 80% of the words in the key, or words semantically similar to the key words (depending on how your semantic space is configured) to receive the "Good1" answer feedback from the tutor.