ESSAY SCORING SYSTEM USING SEMANTIC SIMILARITY APPROACH

Abstract

Essay Scoring System provides a systematic checking of essay based on the similarity
of its meaning to the model as implied on the content. This study automatically score
essays and give feedback to the students about their score. Besides, the system uses
algorithms that analyse the lexical semantics of the words to get the similarity between
the model and student essay which includes the Common Term Frequency (CTF),
Longest Common Subsequence (LCS), and Semantic Distance (SD). The two essay will
undergo the Text Processing Phase which includes the process of Tokenization, Stop
words removal, Stemming and Parts-of-Speech (POS) tagging. It uses the WorldNet
database for word synonymy and semantic references. Word Sense Disambiguation
is also implemented in the study to identify the meaning of the word used in the
context and also to solve the ambiguity of meaning particularly on homonyms. The
scoring follows the predefined criteria for content relevance, spelling, and grammar.
Furthermore, the study conducted tests to the actual users of the system including
teachers. Based on these tests, the computed percentage differences between the teacher
and the system score is 18.03% with an accuracy of 82.18%. The accuracy shows a close
similarity of the system’s score to the teachers’ given score to the essays. Developing the
system faces a challenge in the implementation of the semantic algorithms. Since the study is more capable of evaluating semantic similarity based on word occurrences, it is
best to further the system’s capability of checking the semantic similarity based on the
context of the essay.