The fast growth of aviation communication data has led to a huge increase in the number of Notices to Airmen (NOTAMs). This makes it hard to find and fix mistakes by hand, and mistakes are more likely to happen. Specifically, textual errors resulting from visually or phonetically similar Chinese characters can result in the misinterpretation of crucial aviation data. To tackle this issue, this research suggests a graph-enhanced deep learning framework for the automated correction of NOTAM text. The system uses Relational Graph Convolutional Networks, Convolutional Neural Networks, and a Bidirectional Encoder Representations from Transformers model that improves knowledge to look at how similar characters sound and look. To make a library of similar characters, phonetic and glyph knowledge graphs are made. These graphs are then added to the correction model using multi-hop contrastive learning. During the correction process, characters that are similar to the ones that need to be replaced are taken from the similarity library and evaluated based on the context to find the best replacement. When used on real NOTAM datasets from the Civil Aviation Information Center, the Ctc-CKBERT model greatly improves the accuracy of corrections and the speed of calculations. The proposed method has been shown to work in experiments by lowering character-level errors and making aviation communication systems more reliable.
Dr.N.Sayavathi et al. (Wed,) studied this question.