Presentation Information
[B-14-04]Estimating Commonly Referenced Manual Sections Between Recovery Procedures and Failure Cases
◎Kota Fukuzawa1, Chihiro Sato1, Kengo Funaki1, Naoyuki Tanji1 (1. NTT, Inc.)
Keywords:
operation,recovery procedure,execution decision,graph analysis
AI-based recovery procedure generation for communication network failures has been actively studied. However, inappropriate recovery procedures may cause significant social impacts, making execution decision—verifying the validity of the generated procedures and determining whether they should be carried out—essential. To support this execution decision, AI output verification methods have been proposed that present historical failure cases with failure situations and recovery actions similar to those of the recovery procedure under evaluation. However, existing methods do not explicitly explain why the presented failure cases justify the validity of the recovery procedure. To support execution decisions based on historical failure cases, this paper proposes a method for estimating manual sections that are commonly relevant to both recovery procedures and failure cases.