Hybrid TF-IDF and Character N-Gram-Based Feature Extraction for Ambiguity Detection in Software Requirements

Abstract

Ambiguity in Software Requirement Specifications (SRS) is a critical issue in software engineering. Traditional ambiguity detection approaches rely on lexical or rule-based techniques, which capture surface-level patterns but fail to represent structural variations in requirement statements. This study proposes a hybrid lexical–structural feature extraction framework that integrates TF-IDF with character n-gram representations to capture complementary aspects of ambiguity. The proposed approach is evaluated on the Functional–Non-Functional Requirements (FR–NFR) dataset using Logistic Regression, Random Forest, and Support Vector Machine classifiers under class imbalance conditions. The experimental results demonstrate that single-feature models exhibit unstable precision and low F1-scores, whereas the proposed hybrid representation significantly improves detection performance. The best results are achieved using Support Vector Machine, with an F1-score of 0.6804 and a recall of 0.6429. The findings highlight that ambiguity is a multi-dimensional problem requiring the integration of lexical and structural features. The proposed framework provides a more robust and reliable solution for ambiguity detection in software requirements.