Abstract: Computer-assisted grading in STEM is important and complex task, which would significantly reduce time on task for human graders. Our work sponsored by NSF SBIR # 1721749 investigates various measures of length, its correlation with human scores and achieves accuracy up to 93% when applying a machine learning method to scoring separate sections of lab reports produced by USF students in CHM 2045 and CHM 2046 courses in Spring 2017.