Bocheng Zhou, Jinze Xie, Tiantian Chen and co-authors proposed a language-behavior consistency sensing framework at the Italian National Conference on Sensors in 2026 to detect operational states and potential risks from heterogeneous sensing signals. The paper says existing work has mainly handled either textual information understanding or behavioral data analysis, leaving limited attention to whether textual declarations match executed behaviors. That gap makes some risks hard to detect early when they involve execution deviations but have not yet become significant anomalies.

The framework maps textual sensing signals and behavioral sensing signals into a shared state logic space. It uses a textual state logic extraction module, an observed behavioral state modeling module, and a language-behavior consistency measurement module to compare declared states with executed states and quantify deviations. The experiments used a multisource sensing dataset containing public declaration texts, operation reports, behavioral logs, resource allocation records, and state-response information.

The proposed method achieved the best baseline comparison results, with a language-behavior consistency score (LCS) of 0.742, an AUC of 0.846, an F1-score of 0.811, a Precision of 0.802, and an explanation consistency score (ECS) of 0.821. The paper says it outperformed FinBERT, LSTM, Multimodal Transformer, and the Contrastive Multimodal Model, and could support industrial operation and maintenance, intelligent manufacturing, digital infrastructure management, and other intelligent monitoring scenarios.

Bocheng Zhou, Jinze Xie, Tiantian Chen and co-authors proposed a language-behavior consistency sensing framework at the Italian National Conference on Sensors in 2026 to detect operational states and potential risks from heterogeneous sensing signals. The paper says existing work has mainly handled either textual information understanding or behavioral data analysis, leaving limited attention to whether textual declarations match executed behaviors. That gap makes some risks hard to detect early when they involve execution deviations but have not yet become significant anomalies.

The framework maps textual sensing signals and behavioral sensing signals into a shared state logic space. It uses a textual state logic extraction module, an observed behavioral state modeling module, and a language-behavior consistency measurement module to compare declared states with executed states and quantify deviations. The experiments used a multisource sensing dataset containing public declaration texts, operation reports, behavioral logs, resource allocation records, and state-response information.

The proposed method achieved the best baseline comparison results, with a language-behavior consistency score (LCS) of 0.742, an AUC of 0.846, an F1-score of 0.811, a Precision of 0.802, and an explanation consistency score (ECS) of 0.821. The paper says it outperformed FinBERT, LSTM, Multimodal Transformer, and the Contrastive Multimodal Model, and could support industrial operation and maintenance, intelligent manufacturing, digital infrastructure management, and other intelligent monitoring scenarios.