Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{195980,
        author = {Nikunj C. Gamit and Ajay N. Upadhyaya},
        title = {Negation-Sensitive Calibration for Embedding-Based Automatic Short Answer Grading: A Lightweight Solution},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {906-911},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=195980},
        abstract = {Automatic Short Answer Grading (ASAG) aims to assign scores to short free-text student responses in a way that is consistent with human grading. Early ASAG systems relied heavily on lexical overlap and semantic similarity, while more recent systems use transformer-based encoders and sentence embeddings to compare student answers with reference answers. However, a persistent challenge remains: negation can reverse the meaning of an answer while preserving strong lexical and semantic overlap. As a result, a similarity-based grader may assign a relatively high score to an answer that is actually contradictory. This paper moves beyond diagnostic analysis by proposing a lightweight correction layer, Negation-Sensitive Score Calibration (NSSC), for embedding-based ASAG. NSSC combines sentence-level similarity with concept coverage and a polarity-consistency check over negation cues. We evaluate a standard sentence-embedding baseline, a term-weighted variant, and the proposed NSSC under a negation-oriented protocol in which responses are divided into negation and non-negation subsets. The results show that NSSC can substantially reduce over-scoring on contradictory negated answers while preserving non-negation performance. The paper argues that negation-specific evaluation and lightweight polarity-aware calibration should both be included in ASAG studies because overall metrics can otherwise hide an educationally important failure mode [1], [17], [28]-[30].},
        keywords = {Automatic Short Answer Grading, ASAG, negation, sentence embeddings, negation-sensitive calibration, concept coverage},
        month = {April},
        }