Log-based anomaly detection is a fundamental task for maintaining the reliability and security of modern computing systems. Although deep learning approaches have achieved significant progress, they often incur high computational and memory costs, limiting their applicability in real-time and resource-constrained environments. To address this challenge, we propose LightLog-BERT, an enhanced version of the lightweight LightLog framework that integrates BERT-based contextual embeddings into the log preprocessing pipeline. This integration enriches the semantic representation of log sequences, enabling the model to capture contextual dependencies more effectively while preserving computational efficiency. Experimental evaluations on the HDFS and BGL benchmark datasets show that LightLog-BERT achieves F1-scores of 0.96 and 0.99, respectively, outperforming existing methods and confirming its suitability for real-time and edge-level anomaly detection applications.