Word Ambiguity Identification using POS Tagging in Automatic Essay Scoring

Automatic Essay Scoring (AES) is one of the evaluations carried out by teachers to measure their students. AES is an interesting topic and looking for better accuracy. In AES, the test documents are the student’s answers, and the ground truth is the teacher or book answers. Essay answers from students can consist of one or more sentences. The formation of words in a sentence can affect the meaning of the sentence. In Indonesia, some words can have more than one meaning (polysemy), which can lead to ambiguity of meaning. One example of an Indonesian word with an ambiguous meaning is ‘bisa’. Based on Indonesia Dictionary (KBBI), it has three meanings (capable, toxic substance, and greeting shaman in the traditional ‘pingitan’ ceremony). This research attempts to overcome the ambiguity by integrating POS tagging into the text preprocessing phase before calculating the similarity between student answers and book answers using cosine similarity. The test results show that these approaches can produce an accuracy of up to 79%.