Despite studies of over six decades, research onautomated essay scoring continues to grab ample attentionin  the  Natural  Language  Processing  (NLP)  communityin  part  because  of  its  commercial  and  educational  value.However,  evaluating  such  writing  compositions  or  essaysin   terms   of   reliability   and   time   is   a   very   challengingprocess. The need for reliable and rapid scores has elevatedthe  need  for  a  computer  system  that  can  answer  essayquestions that  fit precise prompts automatically.  NLP andmachine  learning  strategies  use  Automated  Essay  Scoring(AES)  systems  to  solve  the  difficulty  of  scoring  writingtasks.  In  this  paper,  we  suggest  an  AES  approach  thatinvolves   not   only   rule-based   grammar   and   consistencytests,  but  also  the  semantic  similarity  of  sentences,  thusgiving   priority   to   question   prompts.   Similarity   vectorsare used obtained after applying semantic algorithms andcalculated statistical features. Our system uses 22 featureswith  high  predicting  power,  which  is  less  than  currentsystems,  while  considering  every  aspect  a  human  gradermay  focus  on.Predicting  scores  is  achieved  using  the  dataprovided  by  Kaggle’s  ASAP  competition  using  RandomForest.  The  resulting  agreement  between  the  score  of  thehuman grader and the prediction of the system is comparedwith  promising  results  through  experimental  evaluation.