Aerospace Instrument-Making Annotation << Back A Trust-Semantic Method for Predicting Increased Failure Activity of Aviation Subsystems based on BERT Representations of Semi-Structured Information Rytov M.Yu. This paper develops a trust-semantic method for short-term prediction of increased failure activity in aircraft subsystems based on BERT representations of weakly structured operational information. The method combines structured failure-dynamics features, contextual indicators of the operational environment, aggregated BERT representations of defect narratives, and a trust coefficient that downweights incomplete or contradictory messages. Experimental validation was carried out on the open FAA Service Difficulty Reports dataset for 2023–2025; a panel of 24083 observations for 1297 persistent aircraft-model and subsystem pairs was constructed. The developed approach achieved the best top-of-the-list risk prioritization performance among the compared models, namely Precision@20 = 0.9250 and NDCG@20 = 0.9379. These results confirm that trust-aware semantic processing of weakly structured defect reports improves the selection of subsystems that should receive priority preventive actions in reliability management. Keywords: reliability management, aircraft subsystems, weakly structured information, BERT, embeddings, risk-based prioritization, failure activity prediction. DOI: 10.25791/aviakosmos.8.2026.1576 Pp. 26-38.