Research On Guidance Technology Of Audit Methods For Audit Process

Social Security audit is an important supervision measure to the process of socialinsurance. In audit process, auditors need to search variants of information resources, andaudit method is just one class of the most important information resources. As IntelligenceInformation Processing has already been a main research field of computer science,development of audit guidance system, which can easily help auditors search audit methods,will have practical significance.Based on Natural Language Processing and Information Retrieval, according to SocialSecurity Audit, we propose two audit guidance methods: Keywords Search-based AuditMethod Retrieval (KSMR) and Patterns Match-based Audit Question Answer (PMQA).KSMR searches audit methods from existing audit method database using feature informationfrom question descriptions which are given by auditors; PMQA analyses semantic ofquestions which are given by auditors, through matching question descriptions with prestoredsentence patterns in system. When a question matches a sentence pattern, PMQA can get agroup of variable value information, and then uses a template which relates to the sentencepattern to generate answer results with the variable value information.In this thesis, we focus on Social Security Audit and develop the audit guidanceexperiment system AGX. AGX employs140social security audit methods and a group ofsentence patterns summarized from hundreds of social insurance questions. Through theanalysis of experimental result, AGX shows that the proposed two kinds of guidancetechnology can provide audit methods for auditors and have practical value.