Juggling the Jigsaw: Towards Automated Problem Inference from Network Trouble Tickets
summary
NetSieve extracts domain phrases with a compression-backed n-gram pass, filters them with linguistic and information‑theoretic tests, and maps phrases to an ontology to infer problems, activities, actions from 10K+ cloud trouble tickets. It combines statistical NLP, ontology modeling and human-in-the-loop labeling, reports 89%-100% accuracy on labeled data, and surfaces actionable findings: redundancy failures due to faulty cables, version mismatches and misconfigurations; line card/memory/supervisor failures; a new access router half as reliable as its predecessor.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 1:04Network Troubleshooting
- 2:44Questions
- 3:34Ticket Information
- 4:10Challenges
- 5:23Contributions
- 6:27Outline
- 6:49Recall
- 7:21Keywords
- 7:45Clustering
- 8:37Knowledge Building
- 10:12Extract frequently occurring text
- 11:49Identify which of these phrases are important
- 12:59Ontology modeling
- 13:58Ontology classes
- 17:30Accuracy Results
- 18:07Questions Answered
- 19:53Summary
- 20:54Questions Answers
related talks
21mFinding Network Misconfigurations by Automatic Template Inference
24mAutomated Root Cause Analysis
24mNet2Text: Query-Guided Summarization of Network Forwarding Behaviors
24mSergey A. Razin Ph.D., Chief Technology Officer, SIOS Technology @ MLconf SEA
30mShip Reliable RAG: Evals That Grow with Your Retrieval Complexity
36mUsing AI for more automation in failure analysis
11mTransforming Problem-Solving with Data-Driven Insights
19mMulti Agent AI and Network Knowledge Graphs for Change