Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Interpretable sentiment-aware transformer-based model for individual log anomaly detection in distributed systems

Andrés H Catalán1,2,3, Rodrigo A Carrasco4,5, Gonzalo A Ruz1,6,7

  • 1Faculty of Engineering and Sciences, Universidad Adolfo Ibáñez, Santiago, 7941169, Chile.

Scientific Reports
|May 27, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Evolutionary computation for reconstructing threshold networks of the tryptophan operon in Escherichia coli.

Bio Systems·2025
Same author

Oxidative Stress and NLRP3 Inflammasome as Markers of Cardiac Injury Following Cardiopulmonary Bypass: Potential Implications for Patients with Preoperative Heart Failure with Reduced Ejection Fraction.

Antioxidants (Basel, Switzerland)·2025
Same author

Generalization power of threshold Boolean networks.

Bio Systems·2025
Same author

Prostaglandin synthesis mediates the suppression of arcuate Kiss1 neuron activation and pulsatile luteinizing hormone secretion during immune/inflammatory stress in female mice.

Journal of neuroendocrinology·2025
Same author

Brainstem Noradrenergic Neuronal Populations: Dual Effects on Regulating GnRH and LH Secretion.

Endocrinology·2025
Same author

SpectroCVT-Net: A convolutional vision transformer architecture and channel attention for classifying Alzheimer's disease using spectrograms.

Computers in biology and medicine·2024

This study introduces an interpretable transformer model for detecting anomalies in system log entries. The sentiment-aware approach achieves high accuracy and interpretability, crucial for monitoring distributed systems.

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • System Monitoring

Background:

  • Large-scale distributed systems generate vast amounts of log data.
  • Manual log inspection is impractical for anomaly detection.
  • Traditional methods struggle with log volume and variability.

Purpose of the Study:

  • To develop an interpretable, sentiment-aware transformer model for individual log entry anomaly detection.
  • To enhance the transparency of anomaly detection models for system operators.
  • To create a robust and scalable solution for modern distributed infrastructures.

Main Methods:

  • Utilized a Bidirectional Encoder Representations from Transformers with In-Task Pre-Training and Fine-Tuning (BERT-ITPT-FiT) architecture.
  • Incorporated SHapley Additive exPlanations (SHAP) for model interpretability.
Keywords:
Distributed systems monitoringExplainable artificial intelligenceLog anomaly detectionSystem log analysisTransformer-based models

Related Experiment Videos

  • Mapped subword attributions to word-level importance scores for transparency.

    • Main Results:

    • Achieved high F1-scores: up to 99.96% in-domain and 96.97% out-of-domain.
    • Demonstrated strong generalization across different systems.
    • Maintained a processing throughput of ~3,750 log messages per second.

    Conclusions:

    • The proposed model offers a scalable, semantically robust, and interpretable solution for log anomaly detection.
    • Interpretability and semantic robustness are key for reliable anomaly detection in distributed systems.
    • The approach supports large-scale deployment in cloud services and data centers.