The document presents a proposed adaptive approach for subjective answer evaluation. It discusses existing methods that use keyword matching and semantic similarity to evaluate long-form answers. The proposed system extracts keywords from answers, checks for similarity using KNN algorithm while also considering the relationship between keywords and sentences. It further checks for synonyms and similar words using semantic similarity to improve accuracy. Duplication detection is used to compare previous student answers and assign grades based on uniqueness. The system is implemented in ASP.NET and is shown to have better performance than existing methods in features like keywords extraction, synonyms mapping and duplication checking.