Evaluating student knowledge presents a challenge, and one simple and efficient evaluation method is question-answer. Various types of question-answer approaches are available to evaluate a student’s skills. A subjective question is the most relevant method to check all the levels of skills. However, evaluating subjective answers is another challenge, and manually doing this requires an expert who is good at score grading and requires a lot of our time. We propose an automatic subjective answer assessment grading system using a sentence transformer model to solve these difficulties. The proposed method works with two predefined models: the Jaccard approach for checking lexical similarity and the Cosine similarity for semantic similarity. And finally, our sentence transformer model is used to compare the contextual similarity. Combining these value-based hybrid techniques yields a promising score based on the fuzzy logic used to give a subjective answer assessment grading, and it outperforms certain baseline approaches.