Design of an NLP-Based Ambiguity Detection Model for Aviation English Requirements

With the rapid advancement of artificial intelligence (AI) technology, natural language processing (NLP) has been widely applied in the aviation industry. In the requirements analysis of aviation, traditional manual inspection of ambiguity in requirement documents suffers from inefficiency, high costs, and low accuracy. Additionally, iterative updates of requirements further increase the challenges in ambiguity detection. To address these issues, this study proposes a semi-automated ambiguity detection tool that combines machine learning with rule-based methods, aiming to improve detection efficiency and accuracy, reduce economic and labor costs, shorten development cycles, and enhance product quality. Traditional approaches include rule-based methods, which require comprehensive coverage of grammatical structures and standardized sentence patterns (e.g., active voice) to minimize ambiguity, and machine learning-based methods, which rely on corpus training to identify problematic expressions. The hybrid model proposed in this study first constructs a dictionary to screen potentially ambiguous sentences, then employs an NLP pipeline (including modules such as part-of-speech tagging and named entity recognition) to analyze syntactic, semantic, and referential relationships, ultimately determining ambiguity based on a predefined rule set. Experimental results demonstrate that this model achieves an accuracy rate of over 97% in detecting ambiguity in aviation-related English requirement documents, meeting engineering application standards (≥ 95%). Currently, the tool has been successfully implemented in the development of a domestic wide-body aircraft, demonstrating high maturity and providing an efficient and reliable solution for avionics system requirements analysis.